## _wp0415

## Source details

**Canonical URL:** [_wp0415](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0415.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0415.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0415.pdf.json)

---

### Introduction and research questions
- Policymakers and citizens in the second half of the 1990s were increasingly concerned that high external indebtedness in many developing countries is limiting growth and development.
- Pattillo, Poirson, and Ricci (2002) (PPR) found empirical support for a nonlinear impact of debt on growth: at low levels, debt has positive effects on growth; but above particular thresholds or turning points, additional debt begins to have a negative impact on growth.
- Main questions addressed:
  - What are the channels through which debt affects growth?
  - Are nonlinear effects present in the relationship of debt with different sources of growth (physical-capital accumulation, human-capital accumulation, total factor productivity growth)?

### Theoretical channels linking debt to growth
- Capital-accumulation channel:
  - Debt-overhang: large external debt leads investors to lower expectations of returns because of anticipated higher and more distortionary taxes to repay debt, discouraging new domestic and foreign investment and slowing capital-stock accumulation.
  - High indebtedness increases uncertainty about what portion of debt will be serviced, causing investors to hold back.
  - Implication: nonlinear effects of debt on growth likely occur through lower capital accumulation.
- Total factor productivity (TFP) channel:
  - High debt may reduce willingness to undertake difficult and costly policy reforms if benefits will accrue partly to foreign creditors, worsening policy environment and investment efficiency.
  - High uncertainty and instability from debt overhang can hinder incentives to adopt technology or use resources efficiently, causing misallocation toward quick-return activities rather than long-term productivity-enhancing investment.
  - Misallocation and less efficient projects can contribute to slower productivity growth.
- Human-capital channel:
  - High debt may constrain low-income countries' abilities to provide social services, such as education.
  - Human-capital decisions are investment decisions potentially affected by expected high marginal taxes; effects may be hard to detect because human-capital stocks adjust with long lags.

### Empirical approach and estimation methods
- Growth-accounting framework:
  - Combine growth regressions with regressions on sources of growth derived from a consistent growth-accounting exercise; estimate models for physical-capital accumulation, human-capital accumulation, and TFP growth.
- Econometric estimators used:
  - Traditional: Ordinary Least Squares (OLS), instrumental variables (IV), fixed effects (FE).
  - Recent estimators: differenced generalized method of moments (diff-GMM), system GMM (sys-GMM), and identification through heteroskedasticity (IH).
- Nonlinear modeling:
  - Use spline (inverted V) function approach allowing the impact of debt to have a structural break; estimate nonlinear model for growth and its components.
  - Spline specification: low-debt coefficient is γ; high-debt total effect is γ + χ; to translate elasticities for "doubling" debt, multiply coefficients by ln(2).
- Endogeneity and identification:
  - Panel estimators control for unobserved country-specific effects and endogeneity by employing an optimal instrument set based on lagged values of potentially endogenous variables.
  - IH estimator exploits heteroskedasticity across regimes as instruments and is applied as a robustness check on the high-debt subsample.

### Data, sample, and measurement
- Sample and period:
  - Panel regressions for "61 developing countries" spanning Sub-Saharan Africa, Asia, Latin America, and the Middle East over the period "1969–98".
  - Complete dataset: "455 observations for 61 countries over the period 1969–98 (446 observations and 60 countries when using the net present value indicator of debt)."
- Key data sources:
  - Real PPP GDP: WEO database (IMF).
  - Controls: WEO (terms of trade, fiscal balance to GDP, openness), WDI (population growth, secondary education, investment to GDP).
  - Debt measures: Global Development Finance (World Bank) for nominal debt to exports and to GDP; NPV of debt data kindly provided by William Easterly; debt service to exports from GDF.
  - Capital and human capital series: Bosworth and Collins / Susan Collins; schooling from Barro and Lee (2000).
- Growth-accounting and shares:
  - Constant-returns-to-scale production function: Y = A K^α H^β L^(1−α−β).
  - Income shares: physical and human capital income shares α and β are each set equal to "0.33".
  - Human capital adjustment: effective labor H = L*(1 + 0.07*s); in per capita terms h = 1 + 0.07*s, with s = average years of schooling.
- Time aggregation:
  - Three-year averages of variables used to net out short-run fluctuations.
- Key descriptive sample statistics (all values preserved as reported):
  - Per capita GDP growth: Observations "455", Mean "1.1", Std. Dev. "3.8", Between Std. Dev. "2.31", Within Std. Dev. "3.06".
  - Lagged income per capita: Observations "455", Mean "3,562.8", Std. Dev. "2,588.6", Between Std. Dev. "0.80", Within Std. Dev. "0.18".
  - Debt to exports: Observations "455", Mean "283.1", Std. Dev. "330.5", Between Std. Dev. "0.74", Within Std. Dev. "0.53".
  - Debt to GDP: Observations "455", Mean "67.4", Std. Dev. "89.8", Between Std. Dev. "0.69", Within Std. Dev. "0.55".
  - NPV of debt to exports: Observations "446", Mean "234.9", Std. Dev. "218.3", Between Std. Dev. "0.67", Within Std. Dev. "0.51".
  - NPV of debt to GDP: Observations "446", Mean "47.2", Std. Dev. "40.8", Between Std. Dev. "0.65", Within Std. Dev. "0.53".
  - Capital stock growth (per capita): Observations "455", Mean "2.1", Std. Dev. "3.6", Between Std. Dev. "2.60", Within Std. Dev. "2.52".
  - Human capital growth (per capita): Observations "445", Mean "0.6", Std. Dev. "0.5", Between Std. Dev. "0.27", Within Std. Dev. "0.43".
  - TFP growth: Observations "446", Mean "0.2", Std. Dev. "3.2", Between Std. Dev. "1.77", Within Std. Dev. "2.76".
- Country sample includes: "Algeria, Argentina, Bangladesh, Bolivia, Brazil, Cameroon, Chile, China, Colombia, Congo Democratic Republic, Costa Rica, Cote d'Ivoire, Cyprus, Dominican Republic, Ecuador, Egypt, El Salvador, Ethiopia, Ghana, Guatemala, Guyana, Haiti, Honduras, India, Indonesia, Iran, Jamaica, Kenya, Korea, Madagascar, Malawi, Malaysia, Mali, Mauritius, Mexico, Morocco, Mozambique, Myanmar, Nicaragua, Nigeria, Pakistan, Panama, Paraguay, Peru, Philippines, Rwanda, Senegal, Sierra Leone, South Africa, Sri Lanka, Sudan, Tanzania, Thailand, Trinidad and Tobago, Tunisia, Turkey, Uganda, Uruguay, Venezuela, Zambia, and Zimbabwe."

### Estimation details and model specification
- Debt measures included:
  - Four debt stock ratios: nominal debt to exports (DE), nominal debt to GDP (DY), NPV of debt to exports (NE), NPV of debt to GDP (NY).
  - Contemporaneous debt service to exports included to isolate crowding-out effects.
- Threshold selection:
  - Threshold D* chosen using PPR (2002) guidance: "65 percent for debt to exports" and "18 percent for debt to GDP."
- Controls and transformations:
  - Dependent variables: per capita growth (or per capita physical/human capital growth, TFP).
  - Controls: lagged income per capita, investment rate, secondary school enrollment rate, population growth (all in logs), openness (exports plus imports over GDP), fiscal balance, terms of trade growth.
  - Three-year averages used throughout.
  - For growth-accounting, capital and human-capital coefficients are multiplied by "0.33" to obtain their contributions to output growth.
- Inference notes:
  - To translate elasticities for "doubling" debt, coefficients should be multiplied by ln(2).
  - For spline regressions, low-debt coefficient is γ; high-debt total effect is γ + χ.

### Main empirical results and magnitudes
- Nonlinearity and overall magnitude:
  - Impact of debt differs markedly at low vs high levels.
  - At low debt levels: effects generally positive or small and often not significant.
  - At high debt levels: robust, significant negative impact.
    - On average, doubling debt from any initial debt level at or above the threshold reduces per capita growth by about "1 percentage point" per annum.
      - This is derived from an average high-debt coefficient of about "-1.5" multiplied by ln(2).
    - Robustness: IV, diff-GMM, sys-GMM, and IH estimators generally confirm significance; sys-GMM tends to yield smaller magnitudes.
- Channels (sources of growth):
  - High debt has a strong negative effect on:
    - Physical capital accumulation: doubling debt reduces growth in per capita physical capital by almost "1 percentage point" on average.
    - Total factor productivity (TFP): doubling debt reduces TFP growth by almost "1 percentage point" on average.
  - Human capital accumulation: impact of high debt is very small and generally not significant.
  - Decomposition at high debt levels:
    - Approximately one-third of the total negative effect operates via physical-capital accumulation.
    - Approximately two-thirds operates via total factor productivity growth.
- Linear versus nonlinear specifications:
  - Linear models also find negative effects of debt on growth, capital growth, and TFP growth, but magnitudes are on average about "a quarter less" per annum than the high-debt spline estimates — implying linear models understate the high-debt effect when nonlinearity is present.
- Time-series importance and averaging:
  - Decade-averaged estimations yielded similar signs and magnitudes but generally weaker significance; possible reasons include inability to use GMM/IH and that three-year averages capture dynamics better.
- Reverse causality and IH estimates:
  - IH estimation identifies effects of indebtedness on growth and of growth on indebtedness for the high-debt sample.
  - A "1 percentage point" increase in growth reduces debt ratios by about "2–4 percent" (effect larger for nominal debt ratios and for ratios to exports than to income).
  - Normalized comparison: an increase in debt ratios by one standard deviation reduces growth by "0.13–0.26" standard deviations; an increase in output growth by one standard deviation reduces debt ratios by "0.11–0.32" standard deviations.
  - Conclusion: both directions of causality (debt → growth and growth → debt) are significant when accounting for endogeneity.

### Conclusions and policy implications
- Core conclusions:
  - Debt has a nonlinear relationship with growth: at high debt levels doubling debt reduces per capita growth by about "1 percentage point," while at low levels effects are generally positive but often statistically insignificant.
  - The negative high-debt effect operates mainly through reductions in TFP growth (roughly two-thirds) and to a lesser extent through lower physical-capital accumulation (roughly one-third); human-capital effects are negligible in the observed sample and time frame.
  - Results are robust to multiple methods addressing endogeneity (IV, diff-GMM, sys-GMM, IH).
  - Reverse causality from growth to debt is present and quantitatively meaningful, but does not invalidate the debt → growth channel found.
- Policy implications:
  - For the average country in the sample, reducing debt levels should contribute to growth through higher capital accumulation and productivity growth, ceteris paribus.
  - Debt reduction may not yield expected gains if other macroeconomic, structural, or political distortions bind.
  - Relevance to policy debates:
    - Findings inform considerations of the Heavily Indebted Poor Countries (HIPC) Initiative and debt sustainability analyses, but results may not generalize without caution because HIPCs are a nontypical subsample (worse macroeconomic and institutional conditions; continued positive net resource transfers during 1980s and 1990s).
- Suggested avenues for further research:
  - Disaggregate capital channel: does high debt constrain public investment, private investment, or foreign direct investment?
  - Mechanisms behind debt’s negative impact on TFP.
  - Heterogeneity: do effects vary by policy quality or other country-specific factors?
  - Specific analysis for low-income countries and HIPCs.
  - Role of debt flows versus debt stocks: do high past stock with no new borrowing differ from low past stock with high new borrowing?

*Source: _wp0415 - References*

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

### _wp0415 - References

### Introduction and Research Questions
- Policymakers and citizens in the second half of the 1990s were increasingly concerned that high external indebtedness in many developing countries is limiting growth and development.
- Pattillo, Poirson, and Ricci (2002) (PPR) found empirical support for a nonlinear impact of debt on growth: at low levels, debt has positive effects on growth; but above particular thresholds or turning points, additional debt begins to have a negative impact on growth.
- Main questions addressed:
  - What are the channels through which debt affects growth?
  - Are nonlinear effects present in the relationship of debt with different sources of growth (physical-capital accumulation, human-capital accumulation, total factor productivity growth)?

### Theoretical Channels Linking Debt to Growth
- Capital-accumulation channel:
  - Debt-overhang implies that when external debt grows large, investors lower expectations of returns because of anticipation of higher and more distortionary taxes needed to repay debt, discouraging new domestic and foreign investment and slowing capital-stock accumulation.
  - High indebtedness increases uncertainty about what portion of debt will be serviced, causing investors to hold back.
  - These arguments imply nonlinear effects of debt on growth likely occur through lower capital accumulation.
- Total factor productivity (TFP) channel:
  - High debt may reduce willingness to undertake difficult and costly policy reforms if benefits will accrue partly to foreign creditors, worsening policy environment and investment efficiency.
  - High uncertainty and instability from debt overhang can hinder incentives to adopt technology or use resources efficiently, leading to misallocation toward quick-return activities rather than long-term productivity-enhancing investment.
  - Misallocation and less efficient projects can contribute to slower productivity growth.
- Human-capital channel:
  - High debt may constrain low-income countries' abilities to provide social services, such as education.
  - Human-capital decisions are investment decisions potentially affected by expected high marginal taxes; effects may be hard to detect because human-capital stocks adjust with long lags.

### Empirical Approach and Methods
- Growth-accounting framework: combine growth regressions with regressions on sources of growth derived from a consistent growth-accounting exercise; estimate models for physical-capital accumulation, human-capital accumulation, and TFP growth.
- Econometric estimators used:
  - Traditional: Ordinary Least Squares (OLS), instrumental variables, fixed effects.
  - Recent estimators: differenced generalized method of moments (diff-GMM), system GMM, and identification through heteroskedasticity.
- Nonlinear modeling:
  - Utilize spline (inverted V) function approach allowing the impact of debt to have a structural break; estimate nonlinear model for growth and its components.
- Endogeneity and identification:
  - Use panel estimators that control for unobserved country-specific effects and endogeneity by employing an optimal instrument set based on lagged values of potentially endogenous variables.
  - Employ an estimator that uses heteroskedasticity in the data to address endogeneity.

### Key Findings and Quantitative Results
- Negative impact of high debt on growth operates through both:
  - A strong negative effect on physical-capital accumulation.
  - A strong negative effect on total factor productivity growth.
- Magnitudes and thresholds:
  - In PPR, the debt threshold for a negative marginal impact of debt on growth identified at around 65 percent of exports.
  - For countries with high debt levels, doubling debt will reduce output growth by about 1 percentage point and reduce growth in both per capita physical capital and total factor productivity by almost as much.
  - Contributions to the effect of debt on growth: approximately one-third via physical-capital accumulation and two-thirds via total factor productivity growth.
- Nonlinearities by debt level:
  - Debt has, on average, a positive impact on growth and TFP growth at low debt levels, and a negative impact at high debt levels.
  - For physical capital, the average impact of debt at low debt levels is negative, but only about half its average impact at high debt levels.
- Statistical significance and model comparisons:
  - Impact of debt on growth and components at low debt levels is generally not significant (partly due to relatively small number of observations for moderately indebted countries).
  - For highly indebted countries, the impact of high debt on growth is nearly always significant, even after controlling for endogeneity bias.
  - A linear specification that constrains the coefficient of debt to be the same for moderately and highly indebted countries tends to underestimate the negative impact of high debt on growth.
- Time-series importance:
  - Dynamic aspects of debt accumulation matter; the time-series dimension of the panel is crucial in helping to identify the negative impact of high debt on growth.

### Robustness and Causality
- Reverse causality concern: low growth could increase debt as well as high debt lowering growth.
- Strategies to address reverse causality and endogeneity:
  - Use of lagged instruments in panel GMM estimators.
  - Use of an estimator exploiting heteroskedasticity for identification.
- Results are strongly robust to controlling for endogeneity; findings summarized above remain after these controls.

### Paper Organization (preview)
- Remainder of paper organized in five sections: Section II briefly summarizes related literature on debt and growth and the determinants of total factor productivity growth.

*Source: _wp0415 - References*

### Section III describes the data; Section IV describes the estimation methods; Section V

### _wp0415 - Section III describes the data; Section IV describes the estimation methods; Section V

### Theory and related literature
- Debt has potentially nonlinear effects on growth via capital accumulation and total factor productivity (TFP).
- Debt overhang channel: expected future debt-service costs discourage domestic and foreign investment (Krugman, 1988; Sachs, 1989); represented by a “debt Laffer curve.”  
  - PPR (2002) reported the average impact of debt becomes negative at about "160–170 percent of exports or 35–40 percent of GDP."  
  - PPR (2002) found marginal impact becomes negative at about half of these values on average.
- Broader debt overhang interpretation: activities with upfront costs (including policy reforms) are discouraged because future proceeds may be taxed away by creditors, lowering incentives for productivity-enhancing reforms.
- Uncertainty channel: high debt increases macroeconomic uncertainty; investors delay irreversible projects (Serven, 1997), potentially causing misallocation and lower productivity growth.
- Empirical context:
  - Cohen (1997) finds debt becomes excessive at about "50 percent of GDP or 200 percent of exports."
  - Elbadawi et al. (1997) imply growth-maximizing debt to GDP ratio of "97 percent."
  - Reinhart, Rogoff, and Savastano (2003) indicate debt-crisis risk increases at debt levels as low as "15 percent of GNP" for weak-history countries.
- Literature gap: few studies decompose channels (capital vs TFP vs human capital) through which debt affects growth; this paper applies growth accounting to address that gap.

### Data description (Section III)
- Sample and period:
  - Panel regressions for "61 developing countries" spanning Sub-Saharan Africa, Asia, Latin America, and the Middle East over the period "1969–98."
  - Complete dataset: "455 observations for 61 countries over the period 1969–98 (446 observations and 60 countries when using the net present value indicator of debt)."
- Key data sources:
  - Real PPP GDP: WEO database (IMF).
  - Control variables: WEO (terms of trade, fiscal balance to GDP, openness), WDI (population growth, secondary education, investment to GDP).
  - Debt measures: Global Development Finance (World Bank) for nominal debt to exports and to GDP; NPV of debt data kindly provided by William Easterly; debt service to exports from GDF.
  - Capital and human capital series provided by Bosworth and Collins / Susan Collins; schooling from Barro and Lee (2000).
- Growth-accounting specification:
  - Constant-returns-to-scale production function: Y = A K^α H^β L^(1−α−β) (equation (1) and (2) in text).
  - Income shares: physical and human capital income shares α and β are each set equal to "0.33" for the entire sample.
  - Human capital adjustment: effective labor H = L*(1 + 0.07*s) or in per capita terms h = 1 + 0.07*s, with s = average years of schooling.
- Time aggregation and variation:
  - Three-year averages of variables used to net out short-run fluctuations.
  - Descriptive sample statistics (Table 1) include:
    - Per capita GDP growth: Number of observations "455", Mean "1.1", Std. Dev. "3.8", Between Std. Dev. "2.31", Within Std. Dev. "3.06".
    - Lagged income per capita: Observations "455", Mean "3,562.8", Std. Dev. "2,588.6", Between Std. Dev. "0.80", Within Std. Dev. "0.18".
    - Debt to exports: Observations "455", Mean "283.1", Std. Dev. "330.5", Between Std. Dev. "0.74", Within Std. Dev. "0.53".
    - Debt to GDP: Observations "455", Mean "67.4", Std. Dev. "89.8", Between Std. Dev. "0.69", Within Std. Dev. "0.55".
    - NPV of debt to exports: Observations "446", Mean "234.9", Std. Dev. "218.3", Between Std. Dev. "0.67", Within Std. Dev. "0.51".
    - NPV of debt to GDP: Observations "446", Mean "47.2", Std. Dev. "40.8", Between Std. Dev. "0.65", Within Std. Dev. "0.53".
    - Capital stock growth (per capita): Observations "455", Mean "2.1", Std. Dev. "3.6", Between Std. Dev. "2.60", Within Std. Dev. "2.52".
    - Human capital growth (per capita): Observations "445", Mean "0.6", Std. Dev. "0.5", Between Std. Dev. "0.27", Within Std. Dev. "0.43".
    - TFP growth: Observations "446", Mean "0.2", Std. Dev. "3.2", Between Std. Dev. "1.77", Within Std. Dev. "2.76".
- Country list: the sample includes "Algeria, Argentina, Bangladesh, Bolivia, Brazil, Cameroon, Chile, China, Colombia, Congo Democratic Republic, Costa Rica, Cote d'Ivoire, Cyprus, Dominican Republic, Ecuador, Egypt, El Salvador, Ethiopia, Ghana, Guatemala, Guyana, Haiti, Honduras, India, Indonesia, Iran, Jamaica, Kenya, Korea, Madagascar, Malawi, Malaysia, Mali, Mauritius, Mexico, Morocco, Mozambique, Myanmar, Nicaragua, Nigeria, Pakistan, Panama, Paraguay, Peru, Philippines, Rwanda, Senegal, Sierra Leone, South Africa, Sri Lanka, Sudan, Tanzania, Thailand, Trinidad and Tobago, Tunisia, Turkey, Uganda, Uruguay, Venezuela, Zambia, and Zimbabwe."

### Estimation methodology and model specification (Section IV)
- Econometric goal:
  - Augment a standard growth model (conditional convergence) with debt variables and test for nonlinear debt-growth relationships using a spline function (equation (5)): y_it = X_it χ + γ D_it + (D_it − D*) Z_it χ + ε_it, where Z = 1 if debt > D*.
  - Decompose growth into sources: physical capital growth, human capital growth, and TFP growth as endogenous dependent variables.
- Debt measures included:
  - Four debt stock ratios: nominal debt to exports (DE), nominal debt to GDP (DY), NPV of debt to exports (NE), NPV of debt to GDP (NY).
  - Contemporaneous debt service to exports included to isolate crowding-out effects.
- Threshold selection:
  - Threshold D* chosen using PPR (2002) guidance: "65 percent for debt to exports" and "18 percent for debt to GDP."
- Estimators employed:
  - OLS.
  - Instrumental variables (IV) — two-stage least squares using lagged values of endogenous regressors and contemporaneous values of others.
  - Fixed effects (FE).
  - Differenced-GMM (diff-GMM).
  - System-GMM (sys-GMM).
  - Identification through Heteroskedasticity (IH) method (Rigobon family) applied as a robustness check on the high-debt subsample, using heteroscedasticity across regimes as instruments.
- Model controls and transformations:
  - Dependent variables: per capita growth (or per capita physical/human capital growth, TFP).
  - Controls: lagged income per capita, investment rate, secondary school enrollment rate, population growth (all in logs), openness (exports plus imports over GDP), fiscal balance, terms of trade growth.
  - Three-year averages used throughout.
  - For growth-accounting, the contribution weights: capital and human capital income shares α = β = "0.33".
- Inference notes:
  - For spline regressions, the low-debt coefficient is γ; high-debt total effect is γ + χ (sum of debt and "debt extra" term).
  - To translate elasticities for "doubling" debt, coefficients should be multiplied by ln(2).
  - For growth-accounting contributions, multiply capital and human-capital coefficients by "0.33" to obtain their contributions to output growth.

### Main empirical results (Section V)
- Nonlinearity and magnitude:
  - Impact of debt differs markedly at low vs high levels.
  - At low debt levels (coefficients in Table 3): effects generally positive or small and often not significant.
  - At high debt levels (coefficients in Table 4): robust, significant negative impact.
    - On average, doubling debt from any initial debt level at or above the threshold reduces per capita growth by about "1 percentage point" per annum.
      - This is derived from an average high-debt coefficient of about "-1.5" multiplied by ln(2).
    - Robustness: IV, diff-GMM, sys-GMM, and IH estimators generally confirm significance; sys-GMM tends to yield smaller magnitudes.
- Channels (sources of growth):
  - High debt has a strong negative effect on:
    - Physical capital accumulation: doubling debt reduces growth in per capita physical capital by almost "1 percentage point" on average.
    - Total factor productivity (TFP): doubling debt reduces TFP growth by almost "1 percentage point" on average.
  - Human capital accumulation: impact of high debt is very small and generally not significant.
  - Decomposition of the growth effect at high debt levels:
    - Approximately one-third of the total negative effect operates via physical-capital accumulation.
    - Approximately two-thirds operates via total factor productivity growth.
- Comparison with linear specification:
  - Linear models (Table 5) also find negative effects of debt on growth, capital growth, and TFP growth, but magnitudes are on average about "a quarter less" per annum than the high-debt spline estimates — implying linear models understate the high-debt effect if nonlinearity is present.
- Decade-average robustness:
  - Estimations with decade-averaged data yielded similar signs and magnitudes but generally weaker significance; possible reasons include methodology limitations (cannot use GMM/IH), importance of time dimension (within-country variation), or that three-year averages capture dynamics better.
- Reverse causality (IH estimates):
  - IH estimation simultaneously identifies the effect of indebtedness on growth and of growth on indebtedness for the high-debt sample.
  - A "1 percentage point" increase in growth reduces debt ratios by about "2–4 percent" (effect larger for nominal debt ratios and for ratios to exports than to income).
  - Normalized comparison: an increase in debt ratios by one standard deviation reduces growth by "0.13–0.26" standard deviations; an increase in output growth by one standard deviation reduces debt ratios by "0.11–0.32" standard deviations.
  - Conclusion: both directions of causality (debt → growth and growth → debt) are significant when accounting for endogeneity.

### Conclusions and policy implications (Section VI)
- Core conclusions:
  - Debt has a nonlinear relationship with growth: at high debt levels doubling debt reduces per capita growth by about "1 percentage point," while at low levels effects are generally positive but often statistically insignificant.
  - The negative high-debt effect operates mainly through reductions in TFP growth (roughly two-thirds) and to a lesser extent through lower physical-capital accumulation (roughly one-third); human-capital effects are negligible in the observed sample and time frame.
  - Results are robust to multiple methods addressing endogeneity (IV, diff-GMM, sys-GMM, IH).
  - Reverse causality from growth to debt is present and quantitatively meaningful, but does not invalidate the debt → growth channel found.
- Policy implications highlighted by the paper:
  - For the average country in the sample, reducing debt levels should contribute to growth through higher capital accumulation and productivity growth, ceteris paribus.
  - Debt reduction may not yield expected gains if other macroeconomic, structural, or political distortions bind.
  - Relevance to policy debates:
    - Findings inform considerations of the Heavily Indebted Poor Countries (HIPC) Initiative and debt sustainability analyses, but results may not generalize without caution because HIPCs are a nontypical subsample (worse macroeconomic and institutional conditions; continued positive net resource transfers during 1980s and 1990s).
- Suggested avenues for further research (as listed in paper):
  - Disaggregate capital channel: does high debt constrain public investment, private investment, or foreign direct investment?
  - Mechanisms behind debt’s negative impact on TFP.
  - Heterogeneity: do effects vary by policy quality or other country-specific factors?
  - Specific analysis for low-income countries and HIPCs.
  - Role of debt flows versus debt stocks: do high past stock with no new borrowing differ from low past stock with high new borrowing?

*Source: Authors’ calculations and text in the provided IMF working paper excerpt.*

### References

### _wp0415 - References

### Panel data and econometrics
- Arellano, Manuel, and Stephen Bond, 1991, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies, Vol. 58, No. 2, pp. 277–97.
- Blundell, Richard, and Stephen Bond, 1998, “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,” Journal of Econometrics, Vol. 87, pp. 115–143.
- Islam, Nazrul, 1995, “Growth Empirics: A Panel Data Approach,” Quarterly Journal of Economics, Vol. 110, No. 4, pp. 1127–70.
- Rigobon, Roberto, 2003, “Identification through Heteroskedasticity,” Review of Economics and Statistics, Vol. 85, No. 4.
- Rigobon, Roberto, and Brian Sack, 2003, “Measuring the Reaction of Monetary Policy to the Stock Market,” Quarterly Journal of Economics, Vol. 118, No. 2, pp. 636-69.

### Growth theory, empirics, and the Solow framework
- Mankiw, N. Gregory, David Romer, and David Weil, 1992, “A Contribution to the Empirics of Economic Growth,” Quarterly Journal of Economics, Vol. 107, No. 2, pp. 407–37.
- Klenow, P., and A. Rodriguez-Clare, 1997, “The Neoclassical Revival in Growth Economics: Has It Gone Too Far?” in National Bureau of Economic Research Macroeconomics Annual 1997, ed. by B. Bernanke and J. Rotemberg (Cambridge, Massachusetts: MIT Press), p. 73-103.
- Hoeffler, Anke E., 2002, “The Augmented Solow Model and the African Growth Debate,” Oxford Bulletin of Economics and Statistics, Vol. 64, No. 2 (May), pp. 135–58.
- Bosworth, Barry, and Susan Collins, 2003, “The Empirics of Growth: An Update,” (unpublished; Brookings Institution).
- Collins, Susan M., and Barry P. Bosworth, 1996, “Economic Growth in East Asia: Accumulation versus Assimilation,” Brookings Papers on Economic Activity, Vol. 0, No. 2, pp. 135–91.
- Fischer, Stanley, 1993, “The Role of Macroeconomic Factors in Growth,” Journal of Monetary Economics, Vol. 32, No. 3 (December), pp. 485–512.
- Sarel, Michael, 1996, “Nonlinear Effects of Inflation on Economic Growth,” Staff Papers, International Monetary Fund, Vol. 43, No.1, pp. 199–215.
- Edwards, Sebastian, 1998, “Openness, Productivity and Growth: What Do We Really Know?” The Economic Journal, Vol. 108 (March), pp. 383–398.
- Easterly, William R., and Ross Levine, 2001, “It’s Not Factor Accumulation: Stylized Facts and Growth Models, World Bank Economic Review, Vol. 15, No. 2, pp. 177–219.

### Debt, debt overhang, and debt relief
- Cohen, D., 1997, “Growth and External Debt: A New Perspective on the African and Latin American Tragedies,” Centre for Economic Policy Research Discussion Paper, No. 1753, pp. 1-17.
- Krugman, Paul, 1988, “Financing vs. Forgiving a Debt Overhang,” Journal of Development Economics, Vol. 29, pp. 253-268.
- Sachs, Jeffrey, 1989, “The Debt Overhang of Developing Countries,” in Debt Stabilization and Development: Essay in Memory of Carlos Diaz Alejandro, ed. by Calvo, A. Guillermo, and others, (Basil Blackwell: Oxford), p. 80-102.
- Easterly, William R., 2002, “How Did Highly Indebted Poor Countries Become Highly Indebted? Reviewing Two Decades of Debt Relief,” World Development, Vol. 30, No 10, pp. 1677-96.
- Easterly, William R., 2001, “Growth Implosions and Debt Explosions: Do Growth Slowdowns Cause Public Debt Crises?” Contributions to Macroeconomics, Vol. 1, No. 1, Article 1, pp.1-24. Available via the Internet at http://www.bepress.com/bejm/contributions/vol1/iss1/artl.
- Birdsall, Nancy, Stijn Claessens, and Ishac Diwan, 2002, “Policy Selectivity Foregone: Debt and Donor Behavior in Africa,” (unpublished; Center for Global Development).
- Birdsall, Nancy, John Williamson, and Brian Deese, 2002, Delivering on Debt Relief: from IMF gold to a new aid architecture (Washington, D.C: Center for Global Development, Institute for International Economics).
- Elbadawi, Ibrahim, Benno Ndulu, and Njuguna Ndung’u, 1997, “Debt Overhang and Economic Growth in Sub-Saharan Africa,” in External Finance for Low-Income Countries, ed. by Iqbal, Zubair and Ravi Kanbur (Washington: IMF Institute), p. 49-76.
- Pattillo, Catherine, Hélène Poirson, and Luca Ricci, 2002, “External Debt and Growth,” IMF Working Paper 02/69 (Washington: International Monetary Fund), pp. 1-47.
- Reinhart, Carmen, Kenneth S. Rogoff, and Miguel A. Savastano, 2003, “Debt Intolerance,” unpublished (Washington: International Monetary Fund), pp. 1-74.

### Finance, trade openness, and technology spillovers
- Beck, Thorsten, Ross Levine, and Norman Loayza, 2000,“Finance and the Sources of Growth,” Journal of Financial Economics, Vol. 58, pp. 261–300.
- Coe, David T., Elhanan Helpman, and Alexander W. Hoffmaister, 1997, “North-South R&D Spillovers,” Economic Journal, Vol. 107, No. 440, pp. 134–49.
- Miller, Stephen M., and Mukti P. Upadhyay, 2000, “The Effects of Openness, Trade Orientation, and Human Capital on Total Factor Productivity,” Journal of Development Economics, Vol. 63, No. 2, pp. 399–423.
- Servén, Luis, 1997, “Uncertainty, Instability, and Irreversible Investment: Theory, Evidence, and Lessons for Africa,” World Bank Policy Research Working Paper No. 1722, pp.1-44.

### Human capital, education, and measurement
- Barro Robert, and J. Lee, 2000, “International Data on Educational Attainment, Updates and Implications,” Working Paper No. 7911:1-36, (Cambridge, Massachusetts: National Bureau of Economic Research).
- Hogan, Vincent, and Roberto Rigobon, 2003, “Using Unobserved Supply Shocks to Estimate the Returns to Education,” mimeo, Sloan School of Management, pp.1-39.
- Nehru, V., and A. Dhareshwar, 1993, “A New Database on Physical Capital Stock: Sources, Methodology, and Results,” Revista Analisis de Economico, Vol. 8, No. 1, pp. 37–59.

### Miscellaneous methodological and empirical contributions
- Barro, Robert J., N. Gregory Mankiw, and Xavier Sala-i-Martin, 1995, “Capital Mobility in Neoclassical Models of Growth,” The American Economic Review, Vol. 85, pp.103–115.
- Chowdhury, Abdur, 2001, “Foreign Debt and Growth in Developing Countries: A Sensitivity and Causality Analysis Using Panel Data,” paper presented at the WIDER Conference on Debt Relief, Helsinki, 17-18 August, pp. 1-35.
- Cohen, D., 1997, “Growth and External Debt: A New Perspective on the African and Latin American Tragedies,” Centre for Economic Policy Research Discussion Paper, No. 1753, pp. 1-17.
- Rigobon, Roberto, and Brian Sack, 2003, “Measuring the Reaction of Monetary Policy to the Stock Market,” Quarterly Journal of Economics, Vol. 118, No. 2, pp. 636-69.

*Source: _wp0415 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0415.pdf_