## Public Debt and Real GDP: Revisiting the Impact

## Source details

**Canonical URL:** [Public Debt and Real GDP: Revisiting the Impact](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022076-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022076-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022076-print-pdf.pdf.json)

---

### Introduction and motivation
- Objective: revisit and empirically assess the relationship between public debt increases and real GDP, focusing on short- to medium-term impacts in the context of recent global developments.
- Context:
  - Average public debt to GDP ratios reported:
    - Average across all countries projected to reach 98.8 percent in 2021 compared to 83.7 percent in 2019.
    - Advanced Economies (AEs) projected to increase from 103.7 to 122.5.
    - Emerging Markets (EMs) projected to increase from 54.8 to 65.1.
    - Low-Income Developing Economies (LIDCs) projected to increase from 44.2 to 48.5 percentage points (figure reference in source).
  - Policy relevance: public debt sustainability analyses are central to IMF lending decisions and program design; understanding debt impacts on real GDP is crucial for forward-looking assessments.

### Data, shock construction, and empirical method
- Data:
  - Main dataset: World Economic Outlook (WEO) vintages covering all countries from 1995 to the present.
  - Two WEO vintages published every year: April and October.
  - Panel: 178 countries over 1995-2020 for the main analysis.
  - Median public debt to GDP ratio in the full sample: 46.12 percent.
  - Median public debt to GDP ratio in the high-initial-debt subsample: 73.98 percent.
- Construction of debt shocks:
  - Debt shocks = realized growth rate of debt-to-GDP minus forecasted growth rate of debt-to-actual-GDP from October WEO of the same year (forecast errors).
  - Median forecast error (debt to GDP ratio deviation) = 1.23 percent.
  - Largest median forecast error = 8 percent in 2008 (GFC).
- Empirical model:
  - Local projections (Jordà, 2005) to trace k = 0,...,5 year impulse responses of real GDP level to debt shocks.
  - Baseline regression: y_{i,t+k} − y_{i,t−1} = c_{i,k} + d_{t,k} + β_k debt_shock_{i,t} + θ_k Z_{i,t} + ε_{it}.
  - Controls (Z_{i,t}): two lags of real GDP growth and two lags of debt shocks.
  - Country and time fixed effects included; standard errors clustered at the country level.
  - Robustness checks: controlling for initial debt level, testing for Nickell-bias (dropping two lags), alternative WEO vintages (t+2 and t+3 realizations).

### Baseline empirical findings
- Average response:
  - A 1 percent unanticipated increase in the debt to GDP ratio leads to a significant -0.01 percent decrease in the real GDP level at the 3-year horizon.
  - Short-run impacts (k=0,1,2) are negative but not statistically significant; effects materialize by year 3.
- Selected regression coefficients (PUBLIC DEBT SHOCK):
  - K=0 = -0.00144 (std .00316)
  - K=1 = -0.00690 (std .00502)
  - K=2 = -0.00838 (std .00525)
  - K=3 = -0.01040 with standard error 0.00557; marked as * (10 percent significance)
  - K=4 = -0.00896 (std .00554)
  - K=5 = -0.00628 (std .00499)
- Sample sizes and fit by horizon:
  - NUMBER OF COUNTRIES: K=0: 179; K=1: 178; K=2: 176; K=3: 173; K=4: 168; K=5: 163.
  - OBSERVATIONS (N): K=0: 2229; K=1: 2052; K=2: 1876; K=3: 1701; K=4: 1530; K=5: 1366.
  - R-squared (푅^2) by horizon: 0.4033, 0.3962, 0.4185, 0.4615, 0.5060, 0.5459 for K=0,...,5 respectively.
- Illustrative example:
  - Applying the median GFC shock (8 percent median forecast error) to the median country (median public debt 46.12 percent) yields an unanticipated 3.69 percentage points increase in public debt to GDP and reduces output level by about -0.08 percent in 3 years.

### Heterogeneity: initial debt level and debt trajectory
- Initial debt level (threshold = median 46.12 percent):
  - High initial debt subsample (median public debt 73.98 percent):
    - A 1 percent unanticipated increase in debt to GDP leads to a significant -0.02 percent decrease in real GDP 3-5 years after the shock.
    - Illustration: median GFC shock (8 percent) → 5.92 percentage points increase in debt to GDP → -0.16 percent reduction in real GDP.
    - Negative effect is persistent in the medium term.
  - Low initial debt subsample:
    - Responses are not statistically significant.
- Debt trajectory (rising vs. declining over preceding five years):
  - Rising trajectory (median public debt 48.7 percent):
    - A 1 percent unanticipated increase in debt to GDP leads to a significant -0.03 percent decrease in real GDP starting 2 years after the shock; effect reaches -0.04 percent at k=5 and persists.
    - Regression highlights (Rising): PUBLIC DEBT SHOCK(T): 0.0008 (k=0), -0.0060 (k=1), -0.0312*** (k=2), -0.0283** (k=3), -0.0268** (k=4), -0.0405*** (k=5).
  - Declining trajectory (median public debt 41.49 percent):
    - Results are not significant; regression highlights show small and mixed coefficients across horizons.

### Heterogeneity: income classification and HIPC
- Income classification:
  - Low-income countries (median public debt 50.03 percent):
    - A 1 percent unanticipated increase in debt to GDP leads to a significant 0.05 percent increase in real GDP 2 years after the shock; response remains significant up to k=4.
    - Regression highlights (Low-Income): PUBLIC DEBT SHOCK(T): 0.0036 (k=0), 0.0179 (k=1), 0.0457*** (k=2), 0.0380*** (k=3), 0.0315** (k=4), -0.0024 (k=5).
  - High-income countries (median public debt 53.65 percent):
    - A 1 percent unanticipated increase in debt to GDP leads to a significant -0.008 percent reduction in real GDP at k=4.
    - Regression highlights (High-Income): PUBLIC DEBT SHOCK(T): -0.0019 (k=0), -0.0026 (k=1), -0.0057 (k=2), -0.0077* (k=3), -0.0079** (k=4), -0.0006 (k=5).
- HIPC Initiative (public external debt):
  - Context: HIPC Initiative introduced in 1996; it has provided US$76 billion in debt relief so far; debt relief was completed for 36 countries.
  - Pre-HIPC subsample (median public debt 65.28 percent) shows generally negative coefficients (notably negative at k=0: -0.0233).
  - Post-HIPC subsample (median public debt 23.97 percent) shows significant positive responses, including:
    - PUBLIC DEBT SHOCK(T): 0.0669* (k=0), 0.1016 (k=1), 0.2414* (k=2), 0.2171* (k=3), 0.1512* (k=4), 0.0306 (k=5).
    - A 1 percent increase in debt to GDP leads to a significant 0.2414* percent increase in real GDP 2 years after the shock (k=2), the largest magnitude observed.

### Robustness checks and limitations
- Robustness approaches and outcomes:
  - Adding two lags of the initial debt to GDP ratio: estimates remain essentially identical (e.g., PUBLIC DEBT SHOCK(T): -0.00250 (k=0), -0.00887* (k=1), -0.0101*** (k=2), -0.0119*** (k=3), -0.00920 (k=4), -0.00539 (k=5)).
  - Dropping two lags of the dependent variable (Nickell-bias check): estimates remain broadly identical (e.g., PUBLIC DEBT SHOCK(T): -0.00140 (k=0), -0.00720 (k=1), -0.00963 (k=2), -0.0114* (k=3), -0.00957 (k=4), -0.00905 (k=5)).
  - Alternative forecast-error constructions using t+2 and t+3 WEO vintages:
    - t+2 results include PUBLIC DEBT SHOCK(T): 0.00072 (k=0), -0.00101 (k=1), -0.00431 (k=2), -0.00652 (k=3), -0.00665* (k=4), -0.00611 (k=5); median public debt to GDP ratio noted as 35.64 percent.
    - t+3 results include PUBLIC DEBT SHOCK(T): 0.00105 (k=0), -0.00006 (k=1), -0.00184 (k=2), -0.00454 (k=3), -0.00664* (k=4), -0.00491 (k=5); median public debt to GDP ratio noted as 35.10 percent.
  - Overall: main findings remain largely identical across specifications.
- Identification caveats:
  - Forecast errors are used as exogenous shocks; deviations from October WEO forecasts are treated as unexpected.
  - Potential reverse causality would require adjustments within October–December of the same year; authors consider this unlikely.
  - Forecast errors can arise from multiple sources (unexpected policy changes, tax revenue surprises, imperfect information), which may also indicate transmission channels beyond the scope of this paper.
  - Dataset of debt shocks runs through 2019; 2020 debt data were not actual at the time.

### Conclusion, implications, and suggested research directions
- Key empirical conclusions:
  - Aggregate sample: real GDP responds negatively to unanticipated increases in the public debt to GDP ratio, with medium-term effects emerging around year 3.
  - Heterogeneity:
    - (i) Negative effects for countries with a high initial debt level.
    - (ii) Negative effects for countries with a rising debt trajectory over the five preceding years.
    - (iii) Positive effects for low-income countries.
    - (iv) Positive effects for countries that have completed the HIPC debt relief initiative.
  - Subgroup magnitudes can be much larger than aggregate estimates, implying canceling-out effects in the aggregate.
- Policy-relevant implications:
  - Fiscal responses that increase public debt can provide short-term support but may partly or fully negate stimulus effects in the medium term through reductions in real GDP level.
  - Countries with high initial debt or rising debt trajectories face larger negative medium-term output effects from unexpected increases in public debt.
  - Low-income countries and HIPC beneficiaries may derive positive medium-term outcomes from additional borrowing; debt relief can change the growth payoff from borrowing.
  - Policy design should weigh short-run stimulus benefits against medium-term debt dynamics and account for country-specific characteristics.
- Suggested directions for future research:
  - Investigate transmission channels of unanticipated public debt increases across subsamples.
  - Examine how the use of additional debt (public investment versus tax cuts or other spending) affects subsequent growth.
  - Study political/governance interactions (e.g., corruption) and microeconomic channels (natural disasters, governance instability, investor composition).
  - Decompose responses by debt specification: concessional debt, external debt, and short/long-term debt.
  - Analyze the role of monetary conditions and differences in interest-rate environments across countries.

*Italic: IMF Working Paper — Public Debt and Real GDP: Revisiting the Impact (excerpts provided from the source content).*

### Introduction ...........................................................................................................

### Introduction

### Context and motivation
- Understanding the effects of debt on macroeconomic variables (i.e., growth, consumption, etc.) is a key policy question with a long history (see for example Domar, 1944; Barro, 1980; Bernheim, 1987).
- An increase in debt can stem from a rise in government expenditures, an increase in public investment, a reduction in tax revenues or other fiscal changes.
- There is little consensus on the direction and magnitude of the impact of increases in public debt on real GDP, yet this impact is essential for public debt sustainability assessments.

### Role in IMF assessments and policy relevance
- Public debt sustainability analyses are a key element in the IMF’s work on member countries.
- These analyses:
  - Play an essential role in IMF lending decisions.
  - Help assess whether the primary balance needed to stabilize debt under both the baseline and realistic shock scenarios is economically and politically feasible.
  - Provide input into the design of Fund programs, including timing and size of financing, policy choices, and the member’s capacity to repay the Fund.
- For forward-looking assessments, it is crucial to understand how changes in public debt are likely to impact real GDP over the short- to medium-term.

### Recent developments and empirical motivation
- The paper revisits the relationship between public debt and GDP, with renewed interest in the context of the COVID-19 pandemic.
- The pandemic led to a significant contraction in the world economy (Aizenman and Ito, 2020).
- Policy responses, including expansionary fiscal policies, have resulted in sharp increases in public debt levels across the globe.
- Reported averages of public debt to GDP ratios:
  - Average across all countries projected to reach 98.8 percent in 2021 compared to 83.7 percent in 2019, before the pandemic.
  - Advanced Economies (AEs) projected to increase from 103.7 to 122.5 over the same period.
  - Emerging Markets (EMs) projected to increase from 54.8 to 65.1.
  - Low-Income Developing Economies (LIDCs) projected to increase from 44.2 to

### Research objective (as stated)
- To revisit and empirically assess the relationship between public debt increases and real GDP, focusing on short- to medium-term impacts in the context of recent global developments.

*wpiea2022076-print-pdf — Introduction (IMF Working Paper).*

### 48.5 percentage points, respectively (Figure 1). While policy responses that increase public debt may be

### Public Debt and Real GDP: Revisiting the Impact

### Literature and theoretical background
- Theoretical literature: high public debt can create a debt overhang that raises default fears and reduces subsequent growth (Myers 1977; Woo and Kumar 2015). Overlapping generations models and endogenous growth frameworks predict negative long-run effects of high public debt on growth (Blanchard 1985; Diamond 1965; Modigliani 1961; Barro, 1990; Saint-Paul, 1992).
- Empirical findings are mixed: Reinhart and Rogoff (2010) report adverse growth effects at high debt levels (roughly above 90 percent), while other studies find no simple universal threshold (Pescatori et al., 2014; Rahman et al., 2019). Results vary with methodology and sample (e.g., Pattillo et al., 2011; Liaqat, 2019; Cecchetti, Mohanty, and Zampolli, 2011; Eberhardt and Presbitero, 2015; Panizza and Presbitero, 2013).
- Identification challenge: difficulty in isolating exogenous public debt shocks uncorrelated with contemporaneous macro shocks motivates use of forecast errors as instruments for unexpected debt movements.

### Data and empirical methodology
- Data:
  - Main dataset: World Economic Outlook (WEO) vintages covering all countries from 1995 to the present.
  - Two WEO vintages published every year: April and October.
  - Sample period for identified debt shocks runs through 2019 (no COVID-19 actual debt shocks included).
  - Panel: 178 countries over 1995-2020 for the main analysis.
  - Median public debt to GDP ratio in the full sample: 46.12 percent.
  - Median public debt to GDP ratio in the high-initial-debt subsample: 73.98 percent.
- Construction of debt shocks:
  - Debt shocks = realized growth rate of debt-to-GDP minus forecasted growth rate of debt-to-actual-GDP from October WEO of the same year (forecast errors).
  - Median forecast error (debt to GDP ratio deviation) = 1.23 percent.
  - Largest median forecast error = 8 percent in 2008 (GFC).
- Empirical model:
  - Local projections (Jordà, 2005) to trace k = 0,...,5 year impulse responses of real GDP level to debt shocks.
  - Baseline regression: y_{i,t+k} − y_{i,t−1} = c_{i,k} + d_{t,k} + β_k debt_shock_{i,t} + θ_k Z_{i,t} + ε_{it}.
  - Controls (Z_{i,t}): two lags of real GDP growth and two lags of debt shocks.
  - Country and time fixed effects included.
  - Standard errors clustered at the country level.
  - Robustness checks include controlling for initial debt level, testing for Nickell-bias (dropping two lags), and using alternative WEO vintages.

### Main empirical results — baseline
- Average response:
  - A 1 percent unanticipated increase in the debt to GDP ratio leads to a significant -0.01 percent decrease in the real GDP level at the 3-year horizon.
  - Short-run impacts (k=0,1,2) are negative but not statistically significant; effects materialize by year 3.
- Regression coefficient highlights (Table 1, selected):
  - PUBLIC DEBT SHOCK (K=3): -0.01040 with standard error 0.00557; marked as * (10 percent significance).
  - PUBLIC DEBT SHOCK coefficients by horizon: K=0 = -0.00144 (std .00316), K=1 = -0.00690 (std .00502), K=2 = -0.00838 (std .00525), K=3 = -0.01040* (std .00557), K=4 = -0.00896 (std .00554), K=5 = -0.00628 (std .00499).
  - Number of countries used by horizon: K=0: 179; K=1: 178; K=2: 176; K=3: 173; K=4: 168; K=5: 163.
  - Observations (N) by horizon: K=0: 2229; K=1: 2052; K=2: 1876; K=3: 1701; K=4: 1530; K=5: 1366.
  - R-squared (푅^2) by horizon: 0.4033, 0.3962, 0.4185, 0.4615, 0.5060, 0.5459 for K=0,...,5 respectively.
- Illustrative examples:
  - Applying the median GFC shock (8 percent median forecast error) to the median country (median public debt 46.12 percent): an unanticipated 3.69 percentage points increase in public debt to GDP reduces output level by about -0.08 percent in 3 years.
    - (Explanation in text: the 8 percent median shock on debt-to-GDP growth is equivalent to a 3.69 percentage point rise in the median debt-to-GDP ratio of 46.12 percent.)
  - Effect interpretation: small in magnitude but persistent reduction in real GDP level over the medium term.

### Heterogeneity by initial debt level and debt trajectory
- Initial debt level (threshold = median 46.12 percent):
  - High initial debt subsample:
    - A 1 percent unanticipated increase in debt to GDP leads to a significant -0.02 percent decrease in real GDP 3-5 years after the shock.
    - Illustration: median GFC shock (8 percent) applied to median public debt in high-debt subsample (73.98 percent) = 5.92 percentage points increase in debt to GDP → -0.16 percent reduction in real GDP.
    - Negative effect is persistent in the medium term.
  - Low initial debt subsample:
    - Responses are not statistically significant.
- Debt trajectory (rising vs. declining over preceding five years):
  - The negative impulse response to an unanticipated increase in public debt is larger when the debt trajectory over the five preceding years is rising.
  - (Paper defines rising/declining trajectory as positive/negative growth in debt-to-GDP over preceding five years.)

### Additional findings, robustness, and caveats
- Subsample differences:
  - Effect becomes positive for low-income countries and countries having received debt relief from the HIPC Initiative (summary statement in text).
- Robustness checks and identification:
  - Constructing shocks using forecast errors addresses endogeneity concerns between public debt and macro variables.
  - Forecast errors assume October WEO projections have absorbed all public information up to that point; deviations are treated as exogenous shocks.
  - Potential reverse causality would require adjustments within the same quarter (October–December), which the paper deems unlikely.
  - Robustness procedures include controlling for initial debt level, checking for Nickell-bias (dropping lags yields identical results), and alternative vintage checks (t+2 and t+3 WEO vintages).
- Limitations noted:
  - Public debt forecast errors can arise from various sources (unexpected policy changes, tax revenue surprises, imperfect information) which could also point to transmission channels — these are beyond the paper’s scope and suggest directions for future research.
  - Data before 2000 are scarce; debt shocks time series illustrated from 2000 onward.
  - Dataset of debt shocks runs through 2019; 2020 debt data were not actual at the time.

### Policy-relevant implications (as presented)
- Fiscal responses that increase public debt can be effective short-term growth supports but may partly or fully negate stimulus effects in the medium term through increases in debt-to-GDP ratios that reduce real GDP level.
- The magnitude and sign of medium-term effects depend on country characteristics:
  - Countries with high initial debt and rising debt trajectories face larger negative medium-term output effects from unexpected increases in public debt.
  - Low-income countries and HIPC debt-relief recipients may experience different (including positive) medium-term outcomes, implying debt relief and country-specific factors matter for the growth impact of public debt.
- Implication for policy design: weigh short-run stimulus benefits against medium-term debt dynamics, particularly for countries with high initial debt or rising debt trajectories.

*Italic: IMF Working Paper — Public Debt and Real GDP: Revisiting the Impact (excerpts provided from the source content).*

### 30.92 and 73.98 percent in the low and high public debt level subsamples, respectively. Estimates based on equation (2).

### Public Debt and Real GDP: Revisiting the Impact

### Public Debt Trajectory
- Definition: a debt trajectory is rising (declining) when countries had a positive (negative) growth in their debt to GDP ratio over the preceding five years.
- Key empirical findings (rising vs. declining trajectory):
  - Rising debt trajectory:
    - A 1 percent unanticipated increase in the debt to GDP ratio leads to a significant -0.03 percent decrease in real GDP starting 2 years after the shock.
    - The effect reaches -0.04 percent 5 years into the shock and does not die out.
    - Median public debt to GDP ratio is 48.7 percent in the rising debt trajectory subsample.
    - Regression highlights (Table 3, Rising Debt Trajectory):
      - PUBLIC DEBT SHOCK(T): 0.0008 (k=0), -0.0060 (k=1), -0.0312*** (k=2), -0.0283** (k=3), -0.0268** (k=4), -0.0405*** (k=5).
      - NUMBER OF COUNTRIES: 160 (k=0) down to 97 (k=5).
      - NUMBER OF OBSERVATIONS: 1023 (k=0) down to 349 (k=5).
      - R-squared (푹^2): 0.5270 (k=0) to 0.6017 (k=5).
  - Declining debt trajectory:
    - Results are not significant in the declining public debt trajectory subsample.
    - Median public debt to GDP ratio is 41.49 percent in the declining debt trajectory subsample.
    - Regression highlights (Table 3, Declining Debt Trajectory):
      - PUBLIC DEBT SHOCK(T): -0.0009 (k=0), -0.0041 (k=1), 0.0067 (k=2), 0.0008 (k=3), -0.0024 (k=4), 0.0098 (k=5).
      - NUMBER OF COUNTRIES: 122 (k=0) down to 49 (k=5).
      - NUMBER OF OBSERVATIONS: 628 (k=0) down to 157 (k=5).
      - R-squared (푹^2): 0.3484 (k=0) to 0.4695 (k=5).
- Interpretation:
  - The effect in rising-debt countries is four times larger in magnitude over the medium-term than the baseline result, indicating countries accumulating debt are more vulnerable to debt shocks.
  - The result relates to Pescatori, Sandri and Simon (2014) on the importance of debt trajectory for future growth prospects.

### Income Classification
- Sample split by World Bank income classification based on GNI per capita (low-, middle-, high-income).
- Key empirical findings:
  - Low-income countries:
    - A 1 percent unanticipated increase in the debt to GDP ratio leads to a significant 0.05 percent increase in real GDP 2 years after the shock.
    - The response remains significant up to the 4-year horizon.
    - Median public debt to GDP ratio is 50.03 percent in the low-income subsample.
    - Regression highlights (Table 4, Low-Income Countries):
      - PUBLIC DEBT SHOCK(T): 0.0036 (k=0), 0.0179 (k=1), 0.0457*** (k=2), 0.0380*** (k=3), 0.0315** (k=4), -0.0024 (k=5).
      - NUMBER OF COUNTRIES: 38 (k=0) to 27 (k=5).
      - NUMBER OF OBSERVATIONS: 273 (k=0) to 112 (k=5).
      - R-squared (푹^2): 0.2798 (k=0) to 0.7165 (k=5).
  - High-income countries:
    - A 1 percent unanticipated increase in the debt to GDP ratio leads to a significant -0.008 percent reduction in real GDP 4 years after the shock.
    - Median public debt to GDP ratio is 53.65 percent in the high-income subsample.
    - Regression highlights (Table 4, High-Income Countries):
      - PUBLIC DEBT SHOCK(T): -0.0019 (k=0), -0.0026 (k=1), -0.0057 (k=2), -0.0077* (k=3), -0.0079** (k=4), -0.0006 (k=5).
      - NUMBER OF COUNTRIES: 55 (k=0) to 49 (k=5).
      - NUMBER OF OBSERVATIONS: 778 (k=0) to 521 (k=5).
      - R-squared (푹^2): 0.5002 (k=0) to 0.6270 (k=5).
- Interpretation:
  - Low-income countries may benefit significantly from debt shocks, in contrast to high-income countries where the response is negative and smaller in magnitude.
  - Results for middle-income countries are reported elsewhere (Annexes).

### HIPC Initiative (public external debt)
- Context: HIPC Initiative introduced in 1996; it has provided US$76 billion in debt relief so far; debt relief was completed for 36 countries.
- Analysis: sample divided pre- vs. post- completion point of HIPC debt relief; focus on public external debt.
- Key empirical findings:
  - Pre-HIPC subsample:
    - PUBLIC DEBT SHOCK(T): -0.0233 (k=0), -0.0099 (k=1), -0.0033 (k=2), -0.0109 (k=3), -0.0153 (k=4), -0.0008 (k=5).
    - NUMBER OF COUNTRIES: 33 for all k (k=0..5).
    - NUMBER OF OBSERVATIONS: 292 for all k (k=0..5).
    - R-squared (푹^2): 0.1832 (k=0) to 0.6157 (k=5).
    - Median public debt to GDP ratio is 65.28 percent in the pre-HIPC subsample.
  - Post-HIPC subsample:
    - A 1 percent increase in the debt to GDP ratio leads to a significant 0.2414* percent increase in real GDP 2 years after the shock (k=2).
    - PUBLIC DEBT SHOCK(T): 0.0669* (k=0), 0.1016 (k=1), 0.2414* (k=2), 0.2171* (k=3), 0.1512* (k=4), 0.0306 (k=5).
    - NUMBER OF COUNTRIES: 36 (k=0..3), 35 (k=4), 34 (k=5).
    - NUMBER OF OBSERVATIONS: 438 (k=0) down to 261 (k=5).
    - R-squared (푹^2): 0.1866 (k=0) to 0.5975 (k=5).
    - Median public debt to GDP ratio is 23.97 percent in the post-HIPC subsample.
- Interpretation:
  - Post-HIPC countries appear to benefit significantly from extra borrowings after recovering from unmanageable debt burdens; the 0.2414* percent (k=2) effect is the largest magnitude observed in the paper.
  - This is consistent with literature showing positive growth impact from participating in the HIPC Initiative.

### Robustness Analysis
- Approaches tested:
  - Adding two lags of the initial debt to GDP ratio as controls.
    - Result: estimates remain essentially identical.
    - Regression highlights (Table 6, Adding Initial Debt Levels):
      - PUBLIC DEBT SHOCK(T): -0.00250 (k=0), -0.00887* (k=1), -0.0101*** (k=2), -0.0119*** (k=3), -0.00920 (k=4), -0.00539 (k=5).
      - NUMBER OF COUNTRIES: 179 (k=0) down to 163 (k=5).
      - NUMBER OF OBSERVATIONS: 2229 (k=0) down to 1366 (k=5).
      - R-squared (푹^2): 0.4093 (k=0) to 0.5546 (k=5).
  - Dropping two lags of the dependent variable (Nickell-bias check).
    - Result: estimates remain broadly identical.
    - Regression highlights (Table 7, Nickell-Bias):
      - PUBLIC DEBT SHOCK(T): -0.00140 (k=0), -0.00720 (k=1), -0.00963 (k=2), -0.0114* (k=3), -0.00957 (k=4), -0.00905 (k=5).
      - NUMBER OF COUNTRIES: 179 (k=0) down to 163 (k=5).
      - NUMBER OF OBSERVATIONS: 2229 (k=0) down to 1366 (k=5).
      - R-squared (푹^2): 0.3563 (k=0) to 0.2171 (k=5).
  - Alternative constructions of forecast error using realized actual data from t+2 and t+3 WEO vintages.
    - Using t+2 actual data (Table 8):
      - PUBLIC DEBT SHOCK(T): 0.00072 (k=0), -0.00101 (k=1), -0.00431 (k=2), -0.00652 (k=3), -0.00665* (k=4), -0.00611 (k=5).
      - NUMBER OF COUNTRIES: 177 (k=0..2) to 163 (k=5).
      - NUMBER OF OBSERVATIONS: 1890 (k=0..2) to 1380 (k=5).
      - R-squared (푹^2): 0.2504 (k=0) to 0.4942 (k=4).
      - Median public debt to GDP ratio is 35.64 percent (figure note).
    - Using t+3 actual data (Table 9):
      - PUBLIC DEBT SHOCK(T): 0.00105 (k=0), -0.00006 (k=1), -0.00184 (k=2), -0.00454 (k=3), -0.00664* (k=4), -0.00491 (k=5).
      - NUMBER OF COUNTRIES: 173 (k=0..4) to 162 (k=5).
      - NUMBER OF OBSERVATIONS: 1710 (k=0..3) to 1375 (k=5).
      - R-squared (푹^2): 0.2545 (k=0) to 0.4963 (k=4).
      - Median public debt to GDP ratio is 35.10 percent (figure note).
- Overall robustness conclusion:
  - Results remain largely identical across specifications: adding lags of initial debt, addressing Nickell-bias, and using alternative WEO vintages for realized data do not materially change the main findings.

*IMF Working Paper: Public Debt and Real GDP: Revisiting the Impact. Estimates based on equation (2).*

### Conclusion

### Conclusion

### Key empirical findings
- Using a novel construction of exogenous shocks in public debt, real GDP responds negatively to unanticipated increases in the public debt to GDP ratio in the aggregate sample.
- The response varies with country characteristics:
  - (i) An unanticipated increase in the public debt to GDP ratio hurts real GDP level for countries with a high initial debt level.
  - (ii) It hurts real GDP level for countries with a rising debt trajectory over the five preceding years.
  - (iii) It boosts real GDP for countries with a low-income level.
  - (iv) It boosts real GDP for countries that have completed the HIPC debt relief initiative.
- The magnitude of the impact is much more prominent in subgroup analyses, indicating potential canceling out effects in the aggregate sample analysis.
- Country-specific characteristics must be taken into account to fully assess the impact of increasing public debt on real GDP.

### Suggested directions for future research
- Investigate the channels through which unanticipated increases in public debt affect real GDP, separately for each subsample with different macroeconomic characteristics.
- Examine how the usage of additional public debt matters:
  - If debt increases are used for public investment, they are likely to increase subsequent growth (Furceri et al., 2018).
  - If debt increases are channeled to tax cuts or other fiscal spending, the impact on real GDP may differ and depend on governance or corruption.
- Study political/governance interactions: Kim et al. (2017) finds that corruption hurts the impact of debt on growth in countries with high corruption.
- Consider microeconomic channels and specific sources of debt shocks: debt stemming from natural disaster, governance instability, debt investors’ portfolio composition, or structural shifts in countries’ ability to carry debt in relation to financial secular developments and reductions in interest rates (Grigorian et al., 2016; Dell’Ariccia et al., 2018).
- Analyze the role of monetary conditions:
  - Interest rates may remain low in advanced economies with independent central banks, while emerging economies or economies without significant monetary autonomy may face different detrimental impacts on output (Bakker et al., 2019).
- Decompose responses by debt specification: consider concessional debt, external debt, and short/long-term debt, since responses may differ by specification.

### Policy implications
- Low-income countries can benefit from an increase in the debt to GDP ratio.
- Lowering initial levels of debt or keeping a declining debt trajectory increases countries’ potential benefit from additional borrowings.
- Participation in the HIPC Initiative increases countries’ potential to gain from extra borrowings; this can be a positive experience to consider for other debt relief initiatives.

*IMF Working Paper No. WP/22/076*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022076-print-pdf.pdf_