## 1. LIC SUB-CATEGORIES

## Source details

**Canonical URL:** [1. LIC SUB-CATEGORIES](https://www.imf.org/-/media/files/publications/wp/2017/wp17220.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17220.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17220.pdf.json)

---

### Executive summary — purpose, key results, and recommendations
- Purpose of LIC DSAs:
  - Guide LIC borrowing and creditor lending consistent with development goals.
  - Improve IMF and World Bank policy advice on debt.
  - Serve as an early warning system for risks to enable timely preventive financing and policy actions.
- Key assessment results:
  - Debt projection errors are skewed towards optimism for most groups of DSAs assessed.
  - The magnitude of errors tends to increase as forecast horizons extend.
  - The degree of optimism has increased over successive vintages for most groups of LIC DSAs assessed.
  - Evidence of a systematic bias towards optimism—most conclusive for public DSAs—that remains after controlling for major economic and other shocks.
- LIC sub-groups with pronounced optimism include (selected):
  - Those in the two highest income quartiles.
  - ‘Frontier’ LICs with prospects for market access.
  - LICs assessed as displaying ‘moderate risks’ of debt distress.
- Program effect:
  - Public debt projections for countries engaged in IMF-supported adjustment programs display greater optimism than those not under an arrangement.
- Main projection drivers contributing to optimistic DSA debt forecasts:
  - Projected ‘residuals’ (exceptional financing, valuation adjustments, components of the balance of payments).
  - Fiscal deficits.
  - Output growth.
- Policy implications and recommendations:
  - Ensure LIC DSAs and underlying macroeconomic projections are methodologically sound, incorporate all available information, and are guided by realistic assumptions informed by countries’ own experiences and peers.
  - Increase transparency and provide more detailed presentations of projection assumptions for external financing flows (e.g., capital account and other components of external residuals).
  - Exercise caution when using DSAs for policy advice regarding reforms and financing—particularly for countries at ‘moderate risk’ of distress and frontier LICs with increasing non-concessional and non-traditional borrowing options.
- Suggested area for future research:
  - Consider whether findings remain evident over longer periods (e.g., beyond 5 years).

### LIC DSA framework, data, and sub-categories
- DSA features:
  - Debt projections: external and public and publicly-guaranteed (PPG) debt projections over 20-year horizons based on macroeconomic projections produced independently.
  - Risk assessment: four ratings — Low Risk; Moderate Risk; High Risk; In Debt Distress.
  - Policy recommendations: advice regarding policies and borrowing to limit risk of distress.
- Quality-of-policies metric:
  - CPIA index (World Bank) used to classify countries into ‘weak’, ‘medium’, or ‘strong’ with three-year moving averages.
- Data and sample scope:
  - Actual DSAs since 2005; IMF WEO; World Bank WDI.
  - Analysis restricted to post-2005 period.
- Disaggregation: LICs grouped into 8 categories and 24 sub-groups (selected category counts preserved):
  - Frontier LICs (14)
  - Fragile States (27)
  - Small States (14)
  - Other Developing (24)
- Notable sample exclusions:
  - Somalia and South Sudan not included (no DSA through August 2015).
  - Liberia and Timor-Leste excluded for data issues.
  - Albania and Angola graduated from LIC list in 2009 (limited DSA vintages available).

### Methodological approach to projection-error measurement
- Sign convention:
  - Negative error = ratio increased by more than projected by the DSA (underestimation / optimism).
  - Positive error = projections overestimated the degree to which a variable increased (pessimism).
- Three methods discussed:
  - Method A (Standard Simple Difference) — inappropriate when structural breaks present (overstates errors).
  - Method B (Adjusted Difference from base year to year 5) — accurate for static revisions; returns -10 pp in Examples 1/2, -16 pp in dynamic revision Example 3.
  - Method C (Percentage Change from base year to year 5) — returns -10 percent for Example 1 and 3, -5 percent for static revision Example 2; influenced by base effects.
- Chosen empirical strategy:
  - Use both Methods B and C for robustness; reported results focus on Methods B and C.
- Additional analyses:
  - Errors by time horizon (1–5 years).
  - Disaggregation by 24 country sub-groups.
  - Decomposition of external and public DSA errors by projection input component.
  - Tests for bias and influence of exogenous shocks.

### External debt projection findings (Methods B and C, 5-year horizon)
- General: DSA projections underestimate the pace of external debt accumulation for most country groupings; errors generally increase with forecast horizon.
- Country groups with largest optimism in external debt (Year-5 values; Method B = percentage points, Method C = percent; Year-5 Rank shown):
  - Frontier: Method B Year-5 = -29.26 (1); Method C Year-5 = -89.16 (1)
  - Highest income quartile: Method B Year-5 = -19.14 (2); Method C Year-5 = -36.95 (2)
  - Small States: Method B Year-5 = -15.99 (3); Method C Year-5 = -20.12 (5)
  - Middle-High: Method B Year-5 = -13.08 (4); Method C Year-5 = -22.05 (3)
  - Low Risk: Method B Year-5 = -11.64 (5); Method C Year-5 = -30.58 (3)
- Country groups with pessimism in external debt (Year-5 values and ranks):
  - In Distress: Method B Year-5 = 32.84 (23); Method C Year-5 = 34.47 (23)
  - High Risk: Method B Year-5 = 7.87 (19); Method C Year-5 = 33.40 (22)
  - Fragile: Method B Year-5 = 13.50 (22); Method C Year-5 = 30.21 (21)
  - Lowest income quartile: Method B Year-5 = 7.88 (20); Method C Year-5 = 15.87 (20)
- Time-vintage finding:
  - DSAs produced during 2010/2011 show faster projected improvements in debt dynamics (more ex post optimism) compared with vintages from 2007/2008 and 2009/2010.

### Public debt projection findings (Methods B and C, 5-year horizon)
- General: Public debt projections displayed optimism for 19 out of 23 country groups in year 5; errors increase with projection horizon.
- Groups with most pronounced optimism in public debt (Year-5 values and ranks):
  - Highest income quartile: Method B Year-5 = -29.87 (1); Method C Year-5 = -62.23 (2)
  - Middle-High: Method B Year-5 = -9.94 (6); Method C Year-5 = -63.87 (1)
  - Moderate Risk: Method B Year-5 = -12.19 (4); Method C Year-5 = -61.93 (3)
  - Strong CPIA: Method B Year-5 = -13.24 (3); Method C Year-5 = -23.61 (13)
- Groups with pessimism in public debt projections (Year-5 values and ranks):
  - In Distress: Method B Year-5 = 24.77 (23); Method C Year-5 = 22.92 (23)
  - High Risk: Method B Year-5 = -0.91 (18); Method C Year-5 = 6.20 (22)
  - Fragile: Method B Year-5 = 2.32 (20); Method C Year-5 = 1.18 (19)
- IMF program implications (Public Debt):
  - Presence of IMF-supported program reduces optimism for external (public and private) debt projections but increases optimism for public debt projections.
  - Public Debt / Method B Year-5: With IMF Program = -6.73 (9); W/O IMF Program = -2.49 (16)
  - Public Debt / Method C Year-5: With IMF Program = -27.79 (10); W/O IMF Program = -19.29 (14)
- Time-vintage finding:
  - Public DSA vintages from 2010/2011 show increasing optimism in ex post errors compared with vintages from 2007/2008 and 2009/2010.

### Composite rankings across country groups
- Composite ranking = average of four individual rankings (Methods B and C for external and public DSAs). Selected composite ranks:
  - Highest income quartile: Composite rank = 1.8
  - Frontier financing and development category: Composite rank = 3.8
  - Middle-High income quartile: Composite rank = 3.8
  - Moderate Risk (DSA): Composite rank = 5.0
  - Small States: Composite rank = 5.3
- Persistent optimism concentrated among groups with greater access to financing (frontier, higher income quartiles, moderate risk).
- Persistent pessimism concentrated among ‘in distress’, ‘high risk’, lowest income quartile, and fragile states.

### Decomposition of errors by DSA input components (5-year horizon)
- External DSA input components (percent of nominal GDP):
  - Current Account Deficit; Net Foreign Direct Investment; Nominal Interest Rate; Real GDP Growth; Price and Exchange Rate; Exceptional Financing; External Residual.
  - External DSA residuals include exceptional financing, changes in gross foreign assets, errors and omissions, and valuation adjustments.
- Public DSA input components (percent of nominal GDP / percentage points):
  - Primary Budget Deficit; Real Interest Rate; Real GDP Growth; Real Exchange Rate; Other Debt-Creating Flows; Public Residual.
  - Other debt-creating flows include privatization receipts, recognition of contingent liabilities, debt relief, etc.
  - Public residuals refer to changes in public debt not explained by other inputs (e.g., valuation of assets).
- Representative group-level Method B contributions (Overall Error and Rank):
  - External — Frontier: Current Account Deficit -16.75; Foreign Direct Investment 13.95; Residual -33.98; Overall Error -29.26 (1)
  - External — Small States: Current Account Deficit 29.44; Foreign Direct Investment -20.18; Residual -16.42; Overall Error -15.99 (3)
  - Public — Frontier: Primary Balance Deficit -4.93; Interest Rate 2.73; Residual -10.19; Overall Error -10.19 (5)
  - Public — Small States: Primary Balance Deficit -6.38; Growth -4.80; Residual -18.57; Overall Error -18.57 (2)
  - Public — In Distress: Primary Balance Deficit 9.44; Interest Rate 8.21; Other Debt Flows 30.18; Residual 24.77; Overall Error 24.77 (23)
- Note on observations:
  - Analysis focuses on 5-year horizon yielding 148 observations for external DSAs and 145 observations for public DSAs; at 7-year horizon observations fall to 53 (external) and 50 (public).
  - For some groups sums of components do not equal overall error due to missing data for 14 DSAs across 8 countries; authors note this does not affect overall findings or rankings.

### Tests for bias and statistical significance
- Methods used:
  - Method 1: Regression Analysis (Timmermann (2006) approach) — regress outturns on previous year’s projections; unbiased projections imply intercept 휷0 = 0 and slope 휷1 = 1.
  - Method 2: Statistical Test (Schuh (2001) approach) — average cumulative projection errors for 1- to 7-year horizons with 90 percent confidence intervals.
- Results:
  - Regression tests (Table 3): external and public debt ratio projections are not unbiased; joint tests reject unbiasedness for both external and public debt in aggregate.
  - Statistical tests:
    - External debt projections: cumulative errors not significantly different from zero at almost all horizons (90 percent confidence interval includes zero); thus no convincing evidence of bias by this method for external debt projections in aggregate.
    - Public debt projections: exhibit significantly negative cumulative errors over most horizons; errors significantly different from zero from the 3- or 4-year to the 7-year horizons, indicating systematic bias towards optimism.
- Annex II (sub-group tests at 5-year horizon):
  - Method B: Null hypothesis can be rejected across all 23 sub-groupings with a p-value of up to 12 percent (0.12); p-values ≤ 0.12 for 21/23 sub-groups; exceptions: frontier LICs and small states.
  - Method C: Null hypothesis can be rejected across all sub-groupings with a p-value of up to 10 percent (0.10); p-values ≤ 0.10 for 22/23 sub-groups; exception: countries with ‘strong’ CPIA scores.

### Role of shocks: commodity/oil price declines and broader shock exclusion experiments
- Commodity and oil price declines (2012 onward) analysis:
  - Commodity price index: Commodity Industrial Inputs Price Index (agricultural raw materials and metals).
  - Oil price index: average of Dated Brent, West Texas Intermediate, and the Dubai Fateh.
  - Counterfactual experiment using fixed-effects regressions indicates:
    - Commodity Price Experiment:
      - Optimism reduced but remains negative.
      - Magnitude of optimism reduced by about 20 percent for the entire set of countries.
      - Magnitude of optimism reduced by almost 40 percent for the subset of commodity exporters.
    - Oil Price Experiment:
      - Projection errors dampened by 25 percent for the entire set of countries.
      - Errors increase in magnitude by 3 percent for the subset of commodity exporters.
  - Conclusion: commodity and oil price declines contributed to projection errors but are insufficient to explain full magnitude of errors and biases.
- Broader shocks across multiple variables — two exclusion methods:
  - Method 1:
    - Define shock event as a 5-year period where at least one of five variables realized as extreme (bottom 10 percent for growth/export/deflators/FDI, top 10 percent for nominal depreciation); excludes ~one third of dataset.
    - Results after excluding Method 1 shock observations:
      - Public debt projection errors remain negative but diminish by about 64 percent.
      - External DSA projection errors invert from negative to positive (pessimism).
    - Sample sizes noted: External Debt — N=61 countries, observations all/Method 1/Method 2 = 148/95/106. Public Debt — N=60 countries, observations all/Method 1/Method 2 = 146/94/106.
  - Method 2:
    - Approximate DSA bound stress test magnitudes (e.g., nominal depreciation defined at 30 percent; other shocks at standard deviations below historical mean); chosen shock magnitude = 2 SDs yields ~28 percent of dataset defined as shocked.
    - Results after excluding Method 2 shock observations:
      - Forecast errors for external and public debt remain negative—bias towards optimism persists.
      - For external debt, forecast errors are smaller once shock-observations are removed.
      - For public debt, forecast errors in the direction of optimism become more pronounced after excluding shocks (attributed to removal of fragile-country observations that were previously pessimistic).

### Fiscal reaction functions, investment–growth nexus, and external residuals — drivers of optimism
- Fiscal reaction function:
  - Fixed-effects regression of primary budget deficit (share of GDP) on lagged public debt ratio with controls.
  - Historical data (column (1), Table 4): coefficient = -0.02 (negative and significant).
    - Interpretation: increase by 1 percentage point of public debt-to-GDP associated with reduction of primary budget deficit-to-GDP by 0.02 percentage points next period.
  - Projection data (column (2), Table 4): coefficient ≈ 3 times larger and statistically significant — projections assume stronger fiscal reaction than historical LIC experience.
  - Comparison literature:
    - Mendoza and Ostry (2008): -0.023 (industrial countries), -0.035 (emerging markets).
    - D’Erasmo, Mendoza and Zhang (2015): -0.08 (United States), -0.01 (developed countries).
  - Conclusion: DSA projections for LICs assume fiscal consolidation responses stronger than LIC historical experience and comparable to advanced-economy responses.
- Investment–growth nexus:
  - Fixed-effects regressions of real GDP growth on lagged private investment, public investment, and FDI (each separately) with controls.
  - Findings:
    - In DSA projections, coefficients on private investment, public investment, and FDI are positive and significant.
    - In historical data, impacts of these investment measures on next-period growth are not significant.
  - Conclusion: DSA projections assume stronger growth dividends from investment (public and private, and FDI) than supported by historical relationships—evidence of optimism.
- External residuals and balance-of-payments (BOP) components:
  - Fixed-effects regressions of 1-year errors for external residuals on lagged BOP components (shares of GDP) indicate:
    - Errors are significantly related to the size of the capital account balance from the previous period.
    - Implication: projections did not fully incorporate capital account information available at the time; such flows become embedded in the DSA residual.
    - Recommendation: clearer distinction in DSA templates of BOP financing sources could improve transparency and consistency.

### Aggregate conclusions, policy implications, and suggested next steps
- Aggregate conclusions:
  - (i) Errors are evident and skewed towards optimism for most DSAs assessed.
  - (ii) Error magnitudes tend to increase with forecast horizon.
  - (iii) Degree of optimism has increased over successive DSA vintages for most LIC DSAs assessed.
  - (iv) Evidence of systematic bias towards optimism—most conclusive for public DSAs—that remains after controlling for major shocks.
  - Limitation: only a fraction of DSAs’ 20-year projection horizons could be tested against outturns.
- Policy and methodological recommendations:
  - Ensure LIC DSAs and underpinning macroeconomic projections are methodologically sound, incorporate all available information, and are guided by realistic assumptions informed by past performance and peers.
  - Improve transparency and detail on assumptions for external financing flows (e.g., capital account and other components of external residuals).
  - Exercise caution when interpreting and using DSA results for policy advice, program design, and lending decisions—particularly for countries at ‘moderate risk’ of distress and frontier LICs with increasing non-concessional and non-traditional borrowing options.
- Suggested future research:
  - Assess whether findings persist over longer periods (e.g., beyond 5 years).

*Source: IMF working paper (wp17220) — excerpts from Executive Summary, Sections I–III, Methods, Results, Annexes, and related figures/tables as provided in the source content.*

### 1. LIC SUB-CATEGORIES ..................................................................................................

### wp17220 - 1. LIC SUB-CATEGORIES ..................................................................................................

### Executive summary
- Purpose of LIC DSAs: guide LIC borrowing and creditor lending consistent with development goals and long-term debt sustainability; improve IMF and World Bank policy advice on debt; serve as an early warning system for risks to enable timely preventive financing and policy actions.
- Key assessment results:
  - Debt projection errors are skewed towards optimism for most groups of DSAs assessed.
  - The magnitude of errors tends to increase as forecast horizons extend.
  - The degree of optimism has increased over successive vintages for most groups of LIC DSAs assessed.
  - Evidence of a systematic bias towards optimism—most conclusive for public DSAs—that remains after controlling for major economic and other shocks.
- LIC sub-groups with pronounced optimism include (not exclusive to): those in the two highest income quartiles, ‘frontier’ LICs with prospects for market access, and those assessed as displaying ‘moderate risks’ of debt distress.
- Public debt projections for countries engaged in IMF-supported adjustment programs display greater optimism than those not under an arrangement.
- Main projection drivers contributing to optimistic DSA debt forecasts:
  - Projected ‘residuals’ (e.g., exceptional financing, valuation adjustments, components of the balance of payments).
  - Fiscal deficits.
  - Output growth.
- Potential causes of projection errors:
  - Optimism regarding the projected pace and depth of fiscal consolidation (fiscal reaction functions) relative to LICs’ own historical experiences and those of emerging and advanced economies.
  - Optimism about the expected growth dividends of investment (public and private) relative to past LIC experiences and those of countries at higher development levels.
  - External debt forecasts could be improved via incorporation of information regarding the capital account of the balance of payments available at the time of projection.
- Policy implications and recommendations (summarized):
  - Ensure LIC DSAs and underlying macroeconomic projections are methodologically sound, incorporate all available information, and are guided by realistic assumptions informed by countries’ own experiences and peers.
  - Increase transparency and provide more detailed presentations of projection assumptions for external financing flows (e.g., capital account and other components of external residuals).
  - Exercise caution when using DSAs for policy advice regarding reforms and financing—particularly for countries at ‘moderate risk’ of distress and frontier LICs with increasing non-concessional and non-traditional borrowing options.
- Suggested area for future research: consider whether findings remain evident over longer periods (e.g., beyond 5 years).

*Source: IMF working paper (wp17220) — Executive Summary and related sections as provided in the source content.*

### Introduction and context
- LIC DSF objectives and background:
  - Introduced to guide borrowing/lending consistent with development goals, improve IMF and World Bank policy advice, and serve as an early warning system.
  - Launched in the context of major debt relief initiatives: Heavily Indebted Poor Countries (HIPC) Initiative (launched in 1996) and the Multilateral Debt Relief Initiative (MDRI, launched in 2005).
  - By end-2016, 36 countries had received relief under these initiatives, amounting to $76 billion in debt service.
- Observed evolution in DSA projections:
  - Example comparison: unweighted average external debt-to-GDP ratios for a balanced sample of 40 DSAs with first projection years of 2008 and 2012 show actual external debt burdens increased much faster than projected by DSAs in 2008; later DSAs (e.g., 2012) relied on more rapid projected improvements in economic performance to tilt medium- and long-term debt dynamics back towards previously-projected trajectories.
- Central research questions addressed:
  - How have LIC DSAs performed in terms of projection accuracy?
  - Are there differences in performance across sub-groupings of LICs?
  - Do errors result from unanticipated shocks or systematic biases inherent to the DSA and underlying economic projections?
  - What are the most significant drivers of errors?

### LIC DSA framework and features relevant to the assessment
- Main DSA features:
  - Debt Projections: external and public and publicly-guaranteed (PPG) debt projections over 20-year horizons based on macroeconomic projections produced independently.
  - Risk Assessment: risk of debt distress based on indicative external and public debt thresholds, calibrated to assessed quality of policies and institutions. Four possible ratings of external debt distress:
    - Low Risk: all debt burden indicators well below thresholds.
    - Moderate Risk: indicators below thresholds under the baseline, but stress tests indicate thresholds breached by simulated shocks.
    - High Risk: one or more indicators breach thresholds under the baseline.
    - In Debt Distress: country already facing repayment difficulties.
  - Policy Recommendations: advice regarding policies and borrowing to limit risk of distress.
- Quality-of-policies metric: CPIA index (World Bank) used to classify countries into ‘weak’, ‘medium’, or ‘strong’ with three-year moving averages to avoid abrupt changes.

### Data, LIC sub-categories, and methodological considerations
- Data sources: actual DSAs produced for LICs since 2005; IMF’s World Economic Outlook (WEO); World Bank’s World Development Indicators (WDI) databases.
- Sample note: analysis restricted to post-2005 period for data comparability and availability.
- LIC heterogeneity acknowledged: size, economic structure, natural endowments, institutional capacity, vulnerability to non-economic shocks, and history of debt relief.
- Disaggregation: LICs grouped into 8 categories and 24 sub-groups (see Annex I in source). Table 1 (excerpted definitions used in analysis):
  - 1. Financing & Development Category
    - Frontier: LICs that have or have the potential for private capital market access (methodology based on IMF definitions).
    - Small States: based on IMF and World Bank definition.
    - Fragile: countries in fragile situations (World Bank and IMF categorizations).
    - Other Developing: all countries not in other three categories.
  - 2. Risk of Debt Distress
    - Low Risk; Moderate Risk; High Risk; In Distress — DSA classification at time projection is undertaken and DSA finalized.
  - 3. CPIA-Based Policy Strength
    - Strong; Medium; Weak — DSA classification.
  - 4. Position in Income Distribution (Quartiles)
    - Highest; Middle-High; Middle-Low; Lowest — based on GNI per capita (World Bank Atlas method).
  - 5. Commodity Exporters
    - Commodity Exporters: >50 percent of export earnings from fuels and primary commodities.
    - Other Countries.
  - 6. HIPC Status
    - Post-Completion Point HIPCs; Other Countries.
  - 7. IMF Arrangement
    - With IMF Program; W/O IMF Program.
  - 8. Projection Years
    - 2007/2008; 2009/2010; 2010/2011 (grouped by DSA first projection year).
- Sample exclusions and notes:
  - Albania and Angola graduated from LIC list in 2009; only DSAs available were issued in 2008 for Albania and 2007 and 2009 for Angola.
  - Somalia and South Sudan not included (no DSA produced through August 2015).
  - Liberia and Timor-Leste excluded because of data availability and other issues.
  - Unidentified data errors may have implications for results.

### Data revisions, structure breaks, and implications for projection error assessment
- Revisions to economic statistics and compilation methods common in LICs, including:
  - Updates of national accounts (e.g., GDP).
  - Changes in definition and coverage of the public sector (e.g., inclusion of state enterprises or sub-national governments).
  - Materialization of contingent liabilities subsequently added to PPG debt stocks.
- Methodological challenge: DSAs produced after such revisions are often updated (historical series and projections), making them inconsistent and not comparable with older vintages.
- Practical limitation: with hundreds of DSA vintages, identifying when structure breaks occurred and their specific nature (GDP vs. debt stock vs. scope changes) was not feasible for the analysis.
- Illustration (Figure 2 in source): three examples of outturns vs. an initial DSA projection for debt-to-GDP ratio at 5-year horizon:
  - Example 1: No Structural Break — True Error: outturn involves an error of -10 percentage points (pp) of the ratio relative to the initial projection.
  - Example 2: Static Revision — a parallel shift increases the ratio by 20 percentage points from the base year onwards; same -10 pp projection error relative to initial projection but with a static revision (e.g., realization of a contingent liability affecting numerator).
  - Example 3: Dynamic Revision — revision increases the ratio by 20 pp from base year and grows at same annual rate as unrevised outturn’s debt ratio over projection horizon (e.g., revision to nominal GDP).
- Implication: data revisions can create static or dynamic structure breaks that complicate assessment of true projection errors.

### Findings on projection accuracy, drivers, and systematic biases
- Aggregate findings (as summarized in Executive Summary):
  - Widespread optimism in debt projections across most LIC groupings.
  - Increasing optimism over successive DSA vintages.
  - Systematic optimism especially pronounced for public DSAs.
- Sub-group heterogeneity:
  - Pronounced optimism among LICs in higher income quartiles, frontier LICs, and those assessed as ‘moderate risk’.
  - Public debt forecast optimism stronger for countries with IMF-supported adjustment programs than for those without.
- Drivers of projection errors:
  - Significant errors in projected residuals for external DSAs (exceptional financing, valuation adjustments, balance of payments components).
  - Fiscal deficit forecasting errors.
  - Output growth forecasting errors.
- Diagnostic findings on potential causes:
  - Fiscal reaction functions used in projections are more optimistic about pace and depth of fiscal consolidation than historical experience of LICs and that of emerging and advanced economies.
  - Investment–growth relationship assumptions may be overly optimistic about growth dividends from investment compared to historical LIC outcomes and outcomes in higher-income countries.
  - Omission or insufficient use of capital account information at projection time reduces external debt forecast accuracy.

### Policy implications and recommendations
- Strengthen methodological rigor and realism in macroeconomic projections that underlie LIC DSAs:
  - Use country-specific historical experiences and peer comparisons to inform fiscal reaction functions and investment–growth dividends.
  - Incorporate available capital account and other external financing flow information explicitly in external debt projections.
- Improve transparency and presentation of projection assumptions:
  - Provide more detailed documentation of assumptions regarding external financing flows and components of projected residuals.
- Exercise caution in policy and financing advice based on DSAs:
  - Especially for countries at ‘moderate risk’ of distress and frontier LICs with rising non-concessional and non-traditional borrowing options.
- Research agenda:
  - Examine whether the identified optimism and projection biases persist over longer horizons (e.g., beyond 5 years).

*Source: IMF working paper (wp17220) — sections I–III, figures, tables, and explanatory text as provided in the source content.*

### 15. To overcome these challenges, alternative assessment methods were developed. The

### 15. To overcome these challenges, alternative assessment methods were developed.

### Mechanics and rationale of projection-error methods
- Sign convention: Negative error = ratio increased by more than projected by the DSA assessed (underestimation / optimism). Positive error = projections overestimated the degree to which a variable increased (pessimism).
- Method A: Standard Method—Simple Difference (projected value vs. outturn)
  - Without any data revision (Ex. 1), forecast errors are -10 pp of the ratio.
  - With a static structural break (Ex. 2), the simple difference would show -30 pp.
  - With a dynamic structural break where the revised ratio grew at the same rate as the initial ratio (Ex. 3), the simple difference would show -36 pp.
  - Conclusion: Method A overstates errors when structural breaks are present.
- Method B: Adjusted Difference (comparison of adjusted difference from the base year to year 5 for actual outturn versus projection)
  - Returns -10 pp for Ex. 1 (unrevised) and Ex. 2 (static revision).
  - Returns -16 pp for Ex. 3 (dynamic revision), overestimating relative to Ex. 1/2 in that case.
  - Accurate for static revisions, not for dynamic data revisions that preserve growth rates.
- Method C: Percentage Change (comparison of percentage changes from the base year to year 5 for actual outturn versus projection)
  - Returns -10 percent for Ex. 1 (unrevised) and Ex. 3 (dynamic revision).
  - Returns -5 percent for Ex. 2 (static revision), underestimating in that case.
  - Influenced by base effects—series with lower base year ratios will show larger percentage errors, all else equal.

### Summary and methodological choice
- Method A is inappropriate when data include structural breaks, which are common in the DSA dataset.
- Methods B and C have complementary strengths and limitations depending on the revision type and post-break behavior of series.
- The study uses both Methods B and C to ensure robustness; results across Methods B and C are consistent in direction and in the relative magnitude and ordering of overall errors across country sub-groupings.
- All reported results focus on Methods B and C to compensate for data revisions or structural breaks.

### Aggregate approach and additional analyses
- Beyond error assessment, the study:
  - Considers errors by time horizon (1-5 years).
  - Disaggregates error results by 24 country sub-groupings.
  - Decomposes external and public DSA errors by projection input component.
  - Analyzes whether results represent testable ‘biases’ for external and public DSAs.
  - Considers whether errors were influenced by exogenous shocks (e.g., commodity price movements or other macroeconomic shocks).
  - Examines theoretical and other drivers of biases—e.g., assumptions on public and private investment dividends, government fiscal reaction functions, and whether all available balance-of-payments information was incorporated.
- Comparisons are made, where possible, against literature for LICs and other economies.

### External Debt: main findings (Methods B and C)
- General: DSA projections underestimate the pace of external debt accumulation for most country groupings; errors generally increase with forecast horizon.
- Country groups with the largest optimism (negative errors) in external debt projections (ranked 1st by Methods B and C for Year 5):
  - Frontier LICs: Method B Year-5 = -29.26 (percentage points); Method C Year-5 = -89.16 (percent). Year-5 Rank (ascending) = 1.
  - Highest income quartile: Method B Year-5 = -19.14 (pp); Method C Year-5 = -36.95 (percent). Year-5 Rank = 2.
  - Small States: Method B Year-5 = -15.99 (pp) (Rank 3); Method C Year-5 = -20.12 (percent) (Rank 5).
  - Middle-High income quartile: Method B Year-5 = -13.08 (pp) (Rank 4); Method C Year-5 = -22.05 (percent) (Rank 3).
  - Low Risk (DSA): Method B Year-5 = -11.64 (pp) (Rank 5); Method C Year-5 = -30.58 (percent) (Rank 3).
- Country groups with pessimism (positive errors) in external debt projections (Year 5, ranks noted):
  - In Distress: Method B Year-5 = 32.84 (pp) (Rank 23); Method C Year-5 = 34.47 (percent) (Rank 23).
  - High Risk: Method B Year-5 = 7.87 (pp) (Rank 19); Method C Year-5 = 33.40 (percent) (Rank 22).
  - Fragile: Method B Year-5 = 13.50 (pp) (Rank 22); Method C Year-5 = 30.21 (percent) (Rank 21).
  - Lowest income quartile: Method B Year-5 = 7.88 (pp) (Rank 20); Method C Year-5 = 15.87 (percent) (Rank 20).
- Selected numeric excerpts from External Debt / Method B (Year-5 values and Year-5 Rank ascending):
  - Frontier: -29.26 (1)
  - Small States: -15.99 (3)
  - Fragile: 13.50 (22)
  - Other Developing: -4.03 (10)
  - Low Risk: -11.64 (5)
  - Moderate Risk: -10.98 (7)
  - High Risk: 7.87 (19)
  - In Distress: 32.84 (23)
  - Strong: -11.26 (6)
  - Middle-High: -13.08 (4)
  - Middle-Low: 6.32 (18)
  - Lowest: 7.88 (20)
  - With IMF Program: 1.52 (17)
  - W/O IMF Program: -6.32 (9)
- Selected numeric excerpts from External Debt / Method C (Year-5 values and Year-5 Rank ascending):
  - Frontier: -89.16 (1)
  - Highest: -36.95 (2)
  - In Distress: 34.47 (23)
  - High Risk: 33.40 (22)
  - Fragile: 30.21 (21)
  - Small States: -20.12 (5)
  - With IMF Program: 6.34 (17)
  - W/O IMF Program: -14.37 (9)
- Time-vintage finding: DSAs produced during 2010/2011 show faster projected improvements in debt dynamics (more ex post optimism) compared with vintages from 2007/2008 and 2009/2010.

### Public Debt: main findings (Methods B and C)
- General: Public debt projections displayed optimism (underestimation of PPG debt accumulation) for 19 out of 23 country groups in year 5; errors increase with projection horizon.
- Groups with most pronounced optimism in public debt (Year 5, ranks):
  - Moderate Risk: Method B Year-5 = -12.19 (pp) (Rank 4); Method C Year-5 = -61.93 (percent) (Rank 3).
  - Highest income quartile: Method B Year-5 = -29.87 (pp) (Rank 1); Method C Year-5 = -62.23 (percent) (Rank 2).
  - Middle-High income quartile: Method B Year-5 = -9.94 (pp) (Rank 6); Method C Year-5 = -63.87 (percent) (Rank 1).
  - Strong policy (CPIA): Method B Year-5 = -13.24 (pp) (Rank 3); Method C Year-5 = -23.61 (percent) (Rank 13).
- Groups with pessimism in public debt projections (Year 5, ranks):
  - In Distress: Method B Year-5 = 24.77 (pp) (Rank 23); Method C Year-5 = 22.92 (percent) (Rank 23).
  - High Risk: Method B Year-5 = -0.91 (pp) (Rank 18); Method C Year-5 = 6.20 (percent) (Rank 22).
  - Fragile: Method B Year-5 = 2.32 (pp) (Rank 20); Method C Year-5 = 1.18 (percent) (Rank 19).
- IMF program implications (Public Debt):
  - For external (public and private) debt projections, presence of an IMF-supported program reduced optimism relative to countries without a program.
  - For public debt projections, presence of IMF arrangements drives increased optimism at nearly all horizons under both Methods B and C.
  - Public Debt / Method B Year-5: With IMF Program = -6.73 (pp) (Rank 9); W/O IMF Program = -2.49 (pp) (Rank 16).
  - Public Debt / Method C Year-5: With IMF Program = -27.79 (percent) (Rank 10); W/O IMF Program = -19.29 (percent) (Rank 14).
- Time-vintage finding: Public DSA vintages from 2010/2011 show increasing optimism in ex post errors compared with vintages from 2007/2008 and 2009/2010.

### Composite comparison and rankings across country groups
- Composite ranking (average of four individual rankings based on Methods B and C for external and public DSAs) highlights:
  - Frontier financing and development category: Composite rank = 3.8 (Method B/C ranks: 1,1,5,8).
  - Highest income quartile: Composite rank = 1.8 (ranks: 2,2,1,2).
  - Middle-High income quartile: Composite rank = 3.8 (ranks: 4,4,6,1).
  - Moderate Risk (DSA): Composite rank = 5.0 (ranks: 7,6,4,3).
  - Small States: Composite rank = 5.3 (ranks: 3,5,2,11).
  - Groups showing persistent optimism include those with greatest potential to access financing among LICs (e.g., frontier, higher income quartiles, moderate risk).
  - Groups showing persistent pessimism include those ‘in distress’, at ‘high risk’, lowest income quartile, and fragile states.

### Decomposition and inputs
- The paper decomposes DSA forecast errors (Methods B and C) by main economic and financial projection components at the 5-year horizon.
- Table 2 (referenced) lists all economic and financial variables used as inputs for nominal external and public DSA debt-to-GDP ratio projections; these input variables are projected via separate processes and then used as DSA inputs.
- Note: Medium-term projections (1 to 5 years) are based on the IMF’s standard financial programming methods and tools.

*Source: Authors’ calculations. Data: IMF DSA Database.*

### 4. Position in Income Distribution (Quartiles)

### 4. Position in Income Distribution (Quartiles)

### Key group indicators (as reported)
- Middle-Low: 18, 18, 22, 20, 19.5
- Risk of Debt Distress: High Risk: 19, 22, 18, 22, 20.3
- Financing and Development Category: Fragile: 22, 21, 20, 19, 20.5
- Position in Income Distribution (Quartiles): Lowest: 20, 20, 21, 21, 20.5
- Risk of Debt Distress: In Distress: 23, 23, 23, 23, 23.0

### External DSA projections — decomposition of forecast errors (5-year horizon)
- Table 2 identifies projection input components used for DSAs (as a percent of nominal GDP):
  - External DSA components: Current Account Deficit; Net Foreign Direct Investment; Nominal Interest Rate; Real GDP Growth; Price and Exchange Rate; Exceptional Financing; External Residual.
  - Public DSA components: Primary Budget Deficit; Real Interest Rate; Real GDP Growth; Real Exchange Rate; Other Debt-Creating Flows (/2); Public Residual (/3).
  - Footnotes:
    - /1) External DSA residuals include exceptional financing (i.e., changes in arrears and debt relief), changes in gross foreign assets, errors and omissions, and valuation adjustments.
    - /2) Other debt-creating flows include privatization receipts, recognition of contingent liabilities, debt relief, etc.
    - /3) Public DSA residuals refer to any changes in the stock of public debt not explained by all other input components—e.g., changes in the valuation of assets held within the debt portfolio, etc.
- Horizon choice:
  - Analysis focuses on a 5-year horizon to yield 148 observations for external DSAs and 145 observations for public DSAs.
  - At a 7-year horizon observations fall to 53 (external) and 50 (public).
- Main findings on external DSAs at 5-year horizon:
  - External DSAs for most LIC groups displayed errors in the direction of optimism (negative values), most often driven by overestimations for ‘external residuals’ and foreign direct investment.
  - Examples of group-level Method B (Adjusted difference in percentage points) contributions to overall error (Overall Error and Overall Error Rank shown):
    - Frontier: Current Account Deficit -16.75; Foreign Direct Investment 13.95; Residual -33.98; Overall Error -29.26; Overall Error Rank 1
    - Small States: Current Account Deficit 29.44; Foreign Direct Investment -20.18; Residual -16.42; Overall Error -15.99; Overall Error Rank 3
    - Fragile: Current Account Deficit 3.68; Foreign Direct Investment -1.38; Residual 11.12; Overall Error 13.50; Overall Error Rank 22
    - Middle-Low: Current Account Deficit -3.21; Foreign Direct Investment 1.65; Residual 0.30; Overall Error 6.32; Overall Error Rank 18
    - Lowest: Current Account Deficit -6.66; Foreign Direct Investment 6.83; Residual 11.33; Overall Error 7.88; Overall Error Rank 20
  - Method C (percentage differences) examples:
    - Frontier: Current Account Deficit -62.7448; Foreign Direct Investment 48.542; Residual -99.61; Overall Error -89.16; Overall Error Rank 1
    - Small States: Current Account Deficit 52.56; Foreign Direct Investment -22.96; Residual -32.31; Overall Error -20.12; Overall Error Rank 5
    - Fragile: Current Account Deficit 49.12; Foreign Direct Investment -5.48; Residual -22.46; Overall Error 30.21; Overall Error Rank 21
    - Middle-Low: Current Account Deficit 4.52; Foreign Direct Investment 1.31; Residual -11.27; Overall Error 8.71; Overall Error Rank 18
    - Lowest: Current Account Deficit 10.55; Foreign Direct Investment -3.51; Residual 7.75; Overall Error 15.87; Overall Error Rank 20
- Additional notes:
  - For some groups, sums of components do not equal overall error due to missing data for 14 DSAs across 8 countries; this does not affect overall findings or rankings.
  - LIC groups with poorest DSA risk ratings and those in the lowest income quartile showed positive errors (pessimism) driven largely by underestimations of projected external residuals.

### Public DSA projections — decomposition of forecast errors (5-year horizon)
- Main projection components for public DSAs (Method B, percentage points) include: Primary Balance Deficit; Interest Rate; Growth; Exchange Rate; Other Debt Flows; Residual.
- Main findings on public DSAs at 5-year horizon:
  - DSAs for most groups displayed negative errors in the direction of optimism, driven most commonly by errors for projected ‘public residuals’, primary budget deficits, and growth.
  - Few groups displayed positive errors (pessimism), driven by pessimism regarding ‘other debt-creating flows’—notably countries rated ‘in debt distress’ and fragile LICs.
  - Method B examples (percentage points):
    - Frontier: Primary Balance Deficit -4.93; Interest Rate 2.73; Residual -10.19; Overall Error -10.19; Overall Error Rank 5
    - Small States: Primary Balance Deficit -6.38; Growth -4.80; Residual -18.57; Overall Error -18.57; Overall Error Rank 2
    - Fragile: Primary Balance Deficit 1.58; Interest Rate 3.72; Residual 2.32; Overall Error 2.32; Overall Error Rank 20
    - Middle-Low: Primary Balance Deficit 2.06; Interest Rate 1.05; Residual 3.68; Overall Error 3.68; Overall Error Rank 22
    - In Distress: Primary Balance Deficit 9.44; Interest Rate 8.21; Other Debt Flows 30.18; Residual 24.77; Overall Error 24.77; Overall Error Rank 23
  - Method C (percent) examples:
    - Frontier: Primary Balance Deficit -17.67; Interest Rate 6.16; Residual -30.40; Overall Error -30.40; Overall Error Rank 8
    - Small States: Primary Balance Deficit -20.50; Growth -4.47; Residual -26.49; Overall Error -26.49; Overall Error Rank 11
    - Fragile: Primary Balance Deficit 2.42; Interest Rate 3.81; Other Debt Flows 19.09; Residual 1.18; Overall Error 1.18; Overall Error Rank 19
    - Middle-High: Primary Balance Deficit -11.23; Interest Rate 3.37; Residual -63.87; Overall Error -63.87; Overall Error Rank 1
    - Lowest: Primary Balance Deficit 6.54; Interest Rate 5.14; Other Debt Flows 20.97; Residual 2.61; Overall Error 2.61; Overall Error Rank 21
- IMF program effect:
  - DSA public debt projection optimism is greater in aggregate for countries with IMF-supported programs than when there is no Fund arrangement.
  - This is partly driven by greater optimism regarding projected improvements in primary balances under IMF-supported programs.
  - Projected residuals were less optimistic under IMF-supported programs than without Fund support.
  - Method C example rows for IMF arrangement:
    - W/O IMF Program: Primary Balance Deficit -1.86; Interest Rate 1.97; Other Debt Flows 10.81; Residual -19.29; Overall Error -19.29; Overall Error Rank 14
    - With IMF Program: Primary Balance Deficit -11.13; Interest Rate 4.93; Other Debt Flows 9.58; Residual -27.79; Overall Error -27.79; Overall Error Rank 10

### Methods for testing bias and statistical significance
- Two complementary methods used:
  - Method 1: Regression Analysis (Timmermann (2006) approach)
    - Regress outturns on previous year’s projections. Unbiased projections imply regression intercept 휷0 = 0 and slope 휷1 = 1.
    - Regression equation: D_i,t = 휷0 + 휷1 D_i,t^p + u_i,t, where D_i,t is actual debt ratio (growth terms) and D_i,t^p is projected one year earlier.
  - Method 2: Statistical Test (Schuh (2001) approach)
    - Compute average cumulative projection errors for 1-year to 7-year horizons and assess a 90 percent confidence interval around the mean; biased projections will show cumulative errors diverging from zero as horizon increases.
- Application:
  - Table 3 regression tests (Methods B and C) indicate external and public debt ratio projections are not unbiased.
  - Joint tests on intercept and slope reject the null hypothesis of unbiasedness for both external and public debt ratio projections in aggregate.
  - Annex II: Tests applied to 23 country sub-groupings at 5-year horizon show evidence of bias and statistical significance for at least one method (B or C) for both public and external DSAs.

### Results from statistical tests across horizons
- Aggregated findings (Figure 13):
  - External debt projections:
    - Display positive cumulative errors (pessimism) on average from the 1-year to the 3- or 4-year horizon, then inflect towards optimism beyond that horizon.
    - In aggregate, cumulative errors are not significantly different from zero (zero lies within the 90 percent confidence interval at almost all horizons); thus no convincing evidence of bias by this method for external debt projections.
  - Public debt projections:
    - Exhibit significantly negative cumulative errors (optimism) over most horizons.
    - Errors are significantly different from zero from the 3- or 4-year to the 7-year horizons (depending on assessment method), indicating a systematic bias towards optimism.

### Aggregate summaries and interpretation
- Summary bullets:
  - Errors for projected external and public debt burdens tended to display a negative sign and lean in the direction of optimism.
  - For external DSAs, optimistic errors were most often driven by projected residuals, foreign investment, and/or GDP growth.
  - For public DSAs, optimistic errors were most often driven by projected residuals, fiscal performance (primary balances), and growth.
  - Regression analyses found a bias towards optimism for both public and external DSA debt projections in aggregate.
  - Statistical tests identified a systematic bias towards optimism only for public DSA projections.

*Source: IMF DSA Database and authors’ calculations (content from wp17220 - 4. Position in Income Distribution (Quartiles)).*

### 45. This section considers the degree to which observed errors and evidence suggesting biases

### Section 45–76: The degree to which observed errors and evidence suggesting biases towards optimism in DSA debt projections may have been influenced by unanticipated shocks

### Commodity and oil price declines (2012 onward)
- Context:
  - The sharp commodity price decline that began in 2012 and shocks to macroeconomic and financial variables are examined as potential drivers of optimistic DSA projection errors (para. 45–46).
  - Countries relying on commodity exports saw revenues fall after 2012, causing debt ratios to increase more than could have been expected (para. 46).
- Data and indices used:
  - Commodity: Commodity Industrial Inputs Price Index (agricultural raw materials and metals price indices) (footnote 21).
  - Oil: average of three spot prices—Dated Brent, West Texas Intermediate, and the Dubai Fateh (footnote 21).
  - Prices taken from IMF WEO publications (April WEO publications used except September 2006 where April lacked prices) (footnote 21).

### Counterfactual experiment: fixed-effects regressions and inferred counterfactuals
- Regression specification (fixed-effects panel, countries indexed by i and time by t) for projected public debt ratio:
  - 푷푫_i,t^p = 휶_i + 휷푷_t^p + 휸푿_i,t^p + 흐_i,t (para. 48).
  - 푷푫_i,t^p is the projected public debt ratio; 푷_t^p the projected commodity/oil price; 푿_i,t^p controls = filtered projected nominal GDP, filtered projected government primary expenditures and inflation; 흐_i,t random error (para. 48).
- Counterfactual projected public debt ratio:
  - 푷푫_i,t^p̂ = 휶̂_i + 휷̂ 푷_t + 휸̂ 푿_i,t^p where 푷_t is the actual commodity/oil price and coefficients/FE from prior regression (para. 50).
- Simplified difference form used to derive counterfactual:
  - 푷푫_i,t^p̂ = 푷푫_i,t^p + 휷̂ (푷_t^p − 푷_t) − 흐̂_i,t (para. 51).
- Focus: 1-year ahead projections given commodity/oil price projections available over a 2-year horizon; first option chosen (fixed-effects controlling for heterogeneity across countries keeping a time series of one-year projections) (footnote 22).

### Counterfactual experiment results (commodity and oil price shocks)
- Regressions: two regressions run—projected public debt ratio on projected commodity and on projected oil prices—for all countries and the subset of commodity exporters; fixed-effects regressions show a negative and significant relationship between projected commodity/oil price and projected public debt ratio (para. 49).
- Counterfactual error comparison (Figure 14 referenced; red = counterfactual errors, blue = actual errors) (para. 52–53).
- Commodity Price Experiment:
  - Optimism reduced but remains negative.
  - Magnitude of optimism would have been reduced by about 20 percent for the entire set of countries (para. 52).
  - Magnitude of optimism would have been reduced by almost 40 percent for the subset of commodity exporters (para. 52).
- Oil Price Experiment:
  - Projection errors dampened by 25 percent for the entire set of countries (para. 52).
  - Errors increase in magnitude by 3 percent for the subset of commodity exporters (para. 52).
- Conclusion:
  - The decline in oil and commodity prices beginning in 2012 contributed to projection errors in the direction of optimism but is insufficient to explain the full magnitude of projection errors and biases (para. 53).

### Broader shocks across multiple variables: definition and methods
- Variables considered (conceptually similar to external DSA bound stress tests): real GDP growth, export growth, US dollar GDP deflators, FDI as share of GDP, and the nominal exchange rate (depreciation) (para. 54; footnote 24).
- Two approaches to identify and remove shock-affected sample periods:
  - Method 1:
    - Defines a shock event as a 5-year period during which at least one of the five variables had an extreme realization in cumulative terms (para. 55).
    - Bottom 10 percent of distribution of outcomes for these variables identified as major shocks (top 10 percent for nominal depreciation) and corresponding 5-year periods excluded.
    - This method excludes approximately one third of the dataset (para. 55).
    - Rationale: 5-year horizon chosen to capture one or more consecutive shocks similar to one major shock rather than short-horizon deviations (footnote continued).
  - Method 2:
    - Approximates DSA bound stress test notion of shocks: nominal exchange rate depreciation defined at 30 percent; real GDP growth, export growth, GDP deflator, and non-debt creating flows shocks defined as standard deviations below the historical mean (para. 56).
    - After simulations, a shock magnitude of two SDs was chosen; at 2 SDs, 28 percent of the dataset would be defined as being subject to a shock (footnote 26).
    - This method maps closer to stress-test magnitudes used in external DSAs (para. 56).

### Results from excluding major shocks (Methods 1 and 2)
- Method 1 results (para. 57):
  - Major shocks represent an important driver of observed optimism for both public and external DSAs.
  - After excluding shock-associated observations:
    - Public debt projection errors remain negative (optimism) but diminish considerably—by about 64 percent (para. 57).
    - For external DSA projections, errors invert from negative to positive (pessimism) (para. 57).
  - Notes on sample sizes (Figure 15 notes): External Debt—N=61 countries, with 148/95/106 observations (all/Method 1/Method 2). Public Debt—N=60 countries, with 146/94/106 observations (all/Method 1/Method 2) (Figure 15 notes).
- Method 2 results (para. 58):
  - Forecast errors for external and public debt remain negative after excluding major shocks—bias towards optimism persists.
  - For external debt, forecast errors are smaller once shock-observations are removed—forecasts during relatively normal periods display less optimism.
  - For public DSAs, forecast errors in the direction of optimism become even more pronounced after excluding shocks.
    - Possible explanation: many shocks identified under Method 2 affected fragile countries, which displayed positive DSA projection errors in the direction of pessimism; removing fragile countries increased the weight of negative errors in remaining sample, amplifying projection optimism (para. 58).

### Fiscal reaction function and optimism in projected fiscal deficits
- Motivation:
  - Fiscal performance, output growth, and external residuals found earlier to be important drivers of errors and bias (para. 60).
  - The section assesses whether forecast assumptions for fiscal deficits are consistent with historical experience and cross-country evidence (para. 60–61).
- Method:
  - Fixed-effects panel estimation regressing the primary budget deficit (share of GDP) on the public debt ratio lagged by one year, using outturns and projections respectively, with controls = filtered nominal GDP and filtered primary expenditures as share of GDP; Hodrick-Prescott filter with smoothing parameter of 100 used to extract cyclical components (paras. 62, 64 footnote 27).
  - Regression equation:
    - 푷푩푫_i,t = 휶_i + 휷 푷푫_i,t−1 + 휸 푿_i,t + 흐_i,t (para. 63).
- Results:
  - Historical data (column (1) Table 4) yield a negative and significant coefficient of -0.02 (para. 64).
    - Interpretation: an increase by 1 percentage point (pp) of the public debt to GDP ratio is associated with a reduction of the primary budget deficit to GDP ratio of 0.02 points in the next period (footnote 28).
  - Projection data (column (2) Table 4) show a coefficient 3 times larger and statistically significant—DSA projections assume a stronger fiscal reaction than observed in the past for LICs (para. 64).
  - Allowing for first-order autocorrelation strengthens the government’s fiscal reaction in both data and projections, but projections remain more optimistic (para. 64 footnote 29).
  - Comparison with literature:
    - Mendoza and Ostry (2008): coefficient -0.023 for industrial countries and -0.035 for emerging markets (para. 65).
    - D’Erasmo, Mendoza and Zhang (2015): -0.08 for the United States and -0.01 for developed countries in general (para. 65).
    - Implication: assumed fiscal reaction in DSA projections for LICs is larger in magnitude than observed reaction functions for industrial and emerging markets, and very close to that estimated for the United States (para. 65).
  - Robustness: repeating analysis using the initial public debt ratio produces similar results (Annex IV, Table B) (para. 65).
- Conclusion:
  - Although historical fiscal reaction performance in LICs is in line with other groups, LIC DSA projections display considerable optimism relative to their own history and to more advanced economies with greater fiscal flexibility (para. 66).

### Investment–growth nexus and optimism in growth projections
- Motivation:
  - Real GDP growth was often overestimated in projections; one potential explanation is projection assumptions about investment-growth transmission (para. 67).
- Method:
  - Fixed-effects regression of real GDP growth on three types of investment each lagged by one year: (i) private investment, (ii) public investment, (iii) FDI (paras. 69–71).
  - Investment measures: private and public capital formation as share of GDP from WEO; FDI as share of GDP from DSAs; growth data from WEO when investment from WEO, otherwise from DSA database (paras. 70–71).
  - Controls: for private and public investment regressions— inflation and government expenditures; for FDI regressions—imports plus exports as share of GDP to control for openness (para. 72).
  - Regression equation:
    - Growth_i,t = 휶_i + 휷 I_i,t−1 + 휸 X_i,t + 흐_i,t (para. 71).
- Results:
  - Evidence of optimism: in DSA projections, the impact of each of the three investment variables on future growth is assumed to be positive and significant, whereas in historical data these impacts are not significant (para. 73; Table 5 referenced).
  - Literature context:
    - Carkovic and Levine (2002): FDI does not exert a robust positive influence on growth; firm-level studies mixed (para. 74).
    - Warner (2014): no evidence of positive impact of public investment on next-year growth (para. 74).
    - Kahn and Reinhart (1989): private investment has a larger impact on growth than public investment (para. 74).
- Conclusion:
  - DSA projections assume a stronger positive transmission from private and public investment (and FDI) to growth than supported by historical relationships—evidence of optimism in projections regarding growth implications of investment (para. 75).

### External residuals and balance of payments components
- Context:
  - Large external residuals identified earlier are important drivers of DSA projection errors and biases; these residuals include exceptional financing (debt relief, reserves depletion, arrears) and financing flows not otherwise identified as an input to the DSA framework (para. 76).
- Note on robustness:
  - Annex IV illustrates similar exercises adjusting for autocorrelation (Table C) and using 5-year averages (Table D) as robustness tests (para. 76).

*Source: IMF working paper (wp17220), excerpts covering paras. 45–76 and associated figures/tables/footnotes as supplied.*

### 77. In this context, while the current account and most sub-components of the financial

### wp17220 - 77. In this context, while the current account and most sub-components of the financial

### Methods
- Regression approach: fixed effects analysis regressing one-year errors for external residuals on subcomponents of the balance of payments, all lagged by one year (variables expressed as a share of GDP).
- Errors computed using data from DSAs; BOP component data sourced from the IMF WEO (April 2016 edition).
- Note on errors: "The errors in residuals are 1-year errors made at different horizons—in particular, the 1-year errors from year 1, year 2, year 3, etc., are included, if BOP data are available for the regression."

### Results on external residuals and BOP components
- Errors are significantly related to the size of the capital account balance from the previous period, indicating projections did not fully incorporate related information available at the time.
- Implication: flows that would fall within the capital account (e.g., capital transfers and the acquisition/disposal of non-produced non-financial assets) are embedded in the DSA residual; clearer distinction in DSA templates of such BOP financing sources could improve transparency and consistency.
- Empirical findings highlight three drivers of projection errors:
  - Optimism relative to historical and cross-country experiences regarding the government’s fiscal reaction.
  - Optimism regarding the potential impact of investment (public and private) on economic performance.
  - Errors in projected external residuals are at least partly driven by failure to incorporate all available information regarding some BOP components.

### Aggregate findings on DSA performance and biases
- Key conclusions summarized:
  - (i) Errors are evident, and skewed towards optimism for most DSAs assessed.
  - (ii) The magnitude of errors tends to increase as forecast horizons extend.
  - (iii) The degree of optimism has increased over successive vintages for most LIC DSAs assessed.
  - (iv) Evidence of a systematic bias towards optimism—most conclusive for public DSAs—that remains even after controlling for major economic and other shocks.
- Limitation noted: only a fraction of DSAs’ 20-year projection horizons could be tested against outturns.

### Sub-group patterns and program effects
- Pronounced optimism observed for certain LIC sub-groups, including:
  - Those in the two highest income quartiles.
  - ‘Frontier’ LICs with prospects for market access.
  - LICs at ‘moderate risk’ of debt distress.
- Countries engaged in IMF-supported adjustment programs display greater optimism in public debt forecasts than those not under an active Fund program.

### Economic factors driving biases
- Significant drivers across most LIC sub-groups:
  - Projected ‘residuals’ (external DSAs: exceptional financing, valuation adjustments, BOP components).
  - Fiscal deficits.
  - Output growth projections.
- Causes identified:
  - (i) Optimism regarding projected pace and depth of fiscal consolidation (fiscal reaction functions) compared to LICs’ own histories and those of emerging and advanced economies.
  - (ii) Optimism regarding expected growth dividends from investment relative to past LIC experiences and higher-income peers.
  - (iii) External debt forecast errors partly due to not incorporating available information on BOP components.

### Policy and methodological recommendations
- Ensure LIC DSAs and underpinning macroeconomic projections:
  - Are methodologically sound.
  - Incorporate all available information.
  - Are guided by realistic assumptions informed by past performance and peer-country experience.
- Improve transparency and detail on assumptions for external financing flows (e.g., capital account and other components of external residuals) to improve forecasts and align DSAs with IMF country frameworks.
- Exercise caution when interpreting and using DSA results for:
  - Policy advice on reforms.
  - Design of adjustment programs.
  - Lending decisions—particularly for countries at ‘moderate risk’ of distress and frontier LICs with increasing non-concessional and non-traditional borrowing options.
- Suggested area for future research: assess whether findings persist over longer periods (e.g., beyond 5 years).

### Annex highlights
- Annex I — Country sub-categories (selected counts preserved exactly as listed):
  - Frontier LICs (14)
  - Fragile States (27)
  - Small States (14)
  - Other Developing (24)
- Annex II — Significance testing notes:
  - Method B: The null hypothesis can be rejected across all 23 sub-groupings with a p-value of up to 12 percent (0.12). P-values are at or below this threshold for 21/23 sub-groupings; exceptions: frontier LICs and small states (61 and 39 observations, respectively).
  - Method C: The null hypothesis can be rejected across all sub-groupings with a p-value of up to 10 percent (0.10). P-values are at or below this threshold for 22/23 sub-groupings; exception: countries with ‘strong’ CPIA scores (68 observations).
  - Note: sub-groups with fewer observations (e.g., frontier LICs, small states, ‘strong’ CPIA) display greater ambiguity in test results; some sub-groups have as many as 200+ observations.
- Annex III — Projection error decomposition:
  - Decompositions presented for both Methods B and C at the 5-year horizon, adjusted to remove 14 observations missing one or more input components; authors note removing these observations does not appreciably alter relative rankings or broader findings.

*Source: Authors’ calculations. Data: IMF LIC DSAs, IMF WEO (April 2016 edition).*

### 7. IMF Arrangement

### 7. IMF Arrangement

### Annex IV — Regressions and Robustness Checks: Overview
- Source: Authors’ calculations. Data: IMF LIC DSAs.
- Controls used are: (i) inflation, (ii) filtered primary expenditures, and (iii) filtered nominal GDP.
- Hodrick-Prescott filter with a smoothing parameter of 100 used to filter series, following the literature on debt sustainability.

### Table A — Commodity/Oil Price Regressions (fixed effects)
- Method: Fixed effects regressions for (a) the set of all countries and (b) the subset of commodity exporters.
- Key finding: The projected public debt ratio and the projected commodity/oil price exhibit a negative and significant relationship in all four cases.

### Table B — Government Fiscal Reaction Function (deficit vs. initial debt stock)
- Alternative independent variable: the initial public debt ratio, computed as the ratio of last year’s debt stock (reported as the debt stock at the end of the year) over current nominal GDP.
- Controls: same as in the main analysis.
- Key findings:
  - Results are similar to the main analysis.
  - Without the autocorrelation coefficient:
    - Projections are more optimistic for the government’s fiscal reaction (i.e., consolidation) compared to the data, while being less optimistic than in the benchmark case.
  - With the autocorrelation coefficient:
    - Projections are more optimistic compared to the data as well, while displaying more optimism than in the benchmark case.

### Table C — Investment–Growth Relationship with Alternative Autocorrelation Coefficients
- Controls: same as in the main analysis.
- Key findings:
  - For private investment, projections display greater optimism than the historical experience.
  - For public investment and FDI, the relationship between investment and growth is not significant, both in the data and in the projections.

### Table D — Investment–Growth Relationship with 5-year Buckets (Carkovic and Levine method)
- Method: Dependent and independent variables averaged over 5-year periods. Periods considered: 2006-2010, 2011-2015, 2016-2020, 2021-2025, 2025-2030 (when available) for the data and the projections.
- Controls: same as in the main analysis.
- Key findings:
  - Results for private investment and FDI are similar to Table C: projections are more optimistic than the historical experience.
  - For public investment, the relationship with growth is not significant, both in the data and in the projections.

*Source: Authors’ calculations. Data: IMF LIC DSAs.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17220.pdf_