## _wp14125

## Source details

**Canonical URL:** [_wp14125](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14125.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14125.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14125.pdf.json)

---

### Introduction and research scope
- Over the past forty years, the Caribbean has suffered more than 250 natural disasters, most of which were storms.
- During this period, natural disasters killed over 12,000 people and affected over 12 million more.
- Estimated damages of US$19.7 billion in 2010 constant dollars; on average 1 percent of the Caribbean GDP is destroyed every year.
- Region is among the most indebted in the world; study focuses on effects of disasters on per capita GDP and on the debt to GDP ratio, explicitly considering the role of official development assistance (ODA) and debt relief.
- Types of disasters studied: storms and floods; disaster intensity: moderate and severe.
- Sample: panel covering 12 Caribbean countries over 40 years, from 1970 to 2009:
  - Antigua and Barbuda, The Bahamas, Barbados, Dominica, Dominican Republic, Grenada, Haiti, Jamaica, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Trinidad and Tobago.

### Data and variables (Section III)
- Data sources: Penn World Tables version 7.0; IMF World Economic Outlook (WEO, 2011); World Bank WDI (2011); OECD; Paris Club; EM-DAT.
- Endogenous variables (benchmark y_{i,t}, all treated as endogenous):
  - Per capita GDP growth_{i,t}
  - Debt to GDP ratio growth_{i,t}
  - Investment share of GDP growth_{i,t}
  - Government share of GDP growth_{i,t}
  - Inflation rate_{i,t}
  - Trade openness growth_{i,t}
  - Financial depth growth_{i,t}
- Exogenous variables (x_{i,t}):
  - Storms_{i,t}
  - Floods_{i,t}
  - Terms of trade_{i,t}
  - World growth_{i,t}
- Additional specifications: include either ODA to GDP growth or a Debt relief dummy in x_{i,t}.
- Debt relief dummy:
  - Sources: Paris Club Agreements, World Bank Global Development Finance, OECD.Stat.
  - Dummy = 1 when debt relief received surpasses 0.01% of GDP.
  - Alternative dummy with threshold at least 1% of GDP also constructed; results similar; 0.01% threshold chosen for relevance to moderate disasters.
  - When multiple sources, the source showing the largest amount in a year is used; Paris Club set to 1 when an agreement exists (Paris Club does not report amounts).
- Disaster intensity measures (EM-DAT):
  - Primary intensity dummy (following Fomby et al.):
    - intensity^k_{i,t} = { 1, if fatalities^k_{i,t} + 0.3 · total affected^k_{i,t} / population_{i,t} > 0.0001; 0, otherwise }
    - Severe disaster index: threshold increased to 1% of total population.
    - Alternative weighting: total affected weight = 1 tested for robustness.
  - Alternative intensity measure using estimated damages:
    - intensity^k_{i,t} = { 1, if damages^k_{i,t} / GDP_{i,t} > 0.0001; 0, otherwise }
  - EM-DAT inclusion criteria: at least one of i) at least 10 people killed, ii) at least 100 people affected, iii) state of emergency declared or a call for international assistance.

### Econometric framework and inference (Section IV)
- Model: fixed effects unbalanced panel VAR with exogenous variables (VARX):
  - y_{i,t} = α_i + Φ_1 y_{i,t−1} + Φ_2 y_{i,t−2} + Θ_0 x_{i,t} + Θ_1 x_{i,t−1} + Θ_2 x_{i,t−2} + ε_{i,t}
- Fixed effects estimation:
  - LSDV estimator with α_i capturing unobserved time-invariant heterogeneity.
  - Recognized Nickell (1981) bias in LSDV for dynamic panels when T_i is small.
- Bias correction and inference:
  - Implement bootstrap bias-corrected (BSBC) estimator following Pesaran and Zhao (1999), Everaert and Pozzi (2007), and Fomby et al.
  - Bootstrap setup: B = 1,000 for bias-correction; Monte Carlo simulation with 30,000 draws used for impulse-response confidence intervals (Hamilton, 1994).
  - Mean response (generalized impulse response): Ψ(L) = Φ(L)^{-1} Θ(L) (Pesaran and Shin, 1998).
- Diagnostics:
  - Stationarity: Im, Pesaran, Shin (IPS) panel unit-root tests reject unit-root in all cases (column 1 of Table 6); individual ADF and PP tests show majority of countries do not exhibit unit-roots.
  - Lag selection: AIC and SBIC suggest p = q = 1; likelihood ratio tests suggest p = q = 3; compromise choice p = q = 2 for all estimations (one-lag VARX used for ECCU subsample due to smaller sample).
  - Granger causality tests: ODA to GDP growth and debt relief dummy are not Granger caused by endogenous variables → treated as exogenous.

### Key empirical findings — output effects
- Immediate and short-run effects:
  - Both storms and floods have a negative effect on growth; severe disasters generate larger drops in output than moderate ones.
  - Moderate storms: on impact GDP growth falls by more than half a percentage point.
  - Moderate floods: on impact GDP growth falls by 0.1 percentage points.
  - Severe storms: immediate effect above 1 percentage point drop in GDP growth.
  - Severe floods: immediate effect close to 3 percentage points drop in GDP growth.
- Recovery dynamics:
  - Small rebound in economic activity:
    - Year 1 after storms.
    - Year 2 after floods (year 3 after severe floods).
  - Rebound is short lived; relapse to negative growth follows after about a year.
  - After the fifth or sixth year effects are small and tend to disappear.
- Cumulative effects:
  - Moderate storms: economy remains about 0.5 percent below the level without the disaster (no full recovery).
  - Moderate floods: cumulative effect slightly positive on output level.
  - Severe storms: recovery in first few years yields economy slightly above non-disaster output level.
  - Severe floods: small rebound and cumulative effect on output is negative.
- Role of public debt in estimation:
  - Excluding debt from the model overestimates the impact of disasters on output; including public debt is crucial.

### Key empirical findings — debt effects
- Floods:
  - Moderate and severe floods increase debt ratio growth rate.
  - Severe floods: significant increase in growth of debt ratio of about 16 percentage points.
  - Increase in debt after severe floods not reflected in pickup of economic activity until third year of recovery.
- Storms (full Caribbean sample):
  - Storms (in particular severe ones) do not increase debt in the full sample; severe storms can be associated with a decline in debt ratio growth.
  - Severe storm: initial decline in debt growth rate is significant and close to 8 percentage points; decline lasts two periods after a moderate storm and three periods after a severe storm.
- Interpretation:
  - The increase in the debt ratio after a disaster is driven both by lower GDP and by an increase in debt levels to finance recovery and reconstruction.
  - Decline in debt after storms partly explained by external debt relief and aid flows; evidence characterized as weak.

### Role of ODA and debt relief (Section V.B)
- Model with ODA:
  - Including ODA as exogenous yields no material difference versus benchmark; consistent with Raddatz (2009): ODA has only a modest and not significant reduction of disaster effects on output.
  - Explanation: ODA aggregates long-term projects with humanitarian aid and debt relief; not all ODA responds to short-term disasters.
- Model with debt relief dummy:
  - Results differ from benchmark but are not statistically significant → treated as weak evidence.
  - Growth: very similar to cases without debt relief included.
  - Debt:
    - Controlling for debt relief, first 3 years after a moderate storm show an initial increase in debt (not significant).
    - Cumulative effect shows increase in debt ratio level of about 3 percent when debt relief included; contrasts with negligible effect when debt relief excluded.
    - After a severe storm, controlling for debt relief still shows a significant decline in debt ratio growth, but smaller (7.6 percentage points) than without debt relief (7.9 percentage points).
    - Flood results remain unaltered: severe flood initial mean response of debt to GDP ratio growth is an increase close to 15 percentage points.
  - Note: exclusion of debt dummy biases disaster estimates on debt to GDP ratio by capturing part of the negative effect of debt relief.

### ECCU-specific results (Section V.C)
- ECCU subsample: 6 countries; only storms analyzed (few floods: three moderate, zero severe); model estimated with p = q = 1.
- Moderate storms in ECCU:
  - Immediate GDP effect: decline of about 0.7 percentage points (not statistically significant).
  - ECCU does not show strong recovery after moderate storms; economies decline further next year.
  - Cumulative impact: borderline significant lower GDP level of about 2 percent two years after a storm.
  - Debt effect: in year of the storm the debt to GDP ratio grows faster by almost 5 percentage points.
    - Cumulative effect: debt to GDP level more than 5 percent higher after 8 years.
  - Interpretation: smaller, more vulnerable ECCU countries experience larger debt increases and limited recovery; increased debt may be used for social assistance with little growth impact.
- Severe storms in ECCU:
  - Immediate GDP effect: decline close to 2 percentage points (borderline significant) with negative cumulative effect due to very small recovery.
  - Debt: severe storms increase debt ratio growth by almost 2 percentage points on impact; cumulative increase in debt of 5 percent by year 8 (results not statistically significant).
  - Contrast with broader Caribbean: broader sample shows severe storms associated with positive cumulative output effect and decline in debt; ECCU shows the opposite.

### Robustness checks
- Estimations run with both BSBC and LSDV; LSDV results very similar to BSBC given average time sample of 31 years: LSDV bias small.
- Alternative disaster intensity measures tested:
  - Using total affected as intensity (Figures A3–A4).
  - Using estimated damages in U.S. dollars to construct intensity dummy (Figures A5–A6).
- Robustness finding: qualitative response at year 0 is robust across intensity measures; cumulative effects depend on the measure used.

### Policy implications and recommendations
- Disaster-prone countries should:
  - expect debt to rise in the aftermath of disasters (especially if they don’t receive foreign aid),
  - attempt to keep sustainable debt levels, and create enough fiscal space to finance their own recovery.
- Other measures to pursue:
  - establishing disaster funds,
  - increasing insurance coverage for the public and private sector,
  - building resilience by integrating climate and disaster risk into policy frameworks at both the macro and micro levels.
- Targeted aid flows:
  - General ODA has limited immediate mitigating effect after disasters.
  - “Targeted aid flows” (debt relief, debt restructuring, humanitarian aid captured in the debt relief dummy) seem to play a role in lessening negative effects on debt after storms; evidence is weak.

### Selected key statistics and data notes
- Average time sample in the data: 31 years.
- Table 3 (selected rows):
  - Per capita GDP growth: Obs.359 Mean2.39 Std. Dev.5.37 Min.−19.82 Max.18.77
  - Debt ratio growth: Obs.359 Mean3.01 Std. Dev.16.08 Min.−100.48 Max.77.77
  - Debt relief dummy: Obs.359 Mean0.32 Std. Dev.0.47 Dummy1115
  - Storms dummy (moderate): Obs.359 Mean0.17 Std. Dev.0.38 Dummy161
  - Floods dummy (moderate): Obs.359 Mean0.07 Std. Dev.0.26 Dummy126
  - Storms dummy (severe): Obs.359 Mean0.05 Std. Dev.0.21 Dummy117
  - Floods dummy (severe): Obs.359 Mean0.01 Std. Dev.0.09 Dummy13
- Table 4 (ECCU selected rows):
  - Per capita GDP growth: Obs.158 Mean3.43 Std. Dev.5.05 Min.−13.26 Max.15.88
  - Debt ratio growth: Obs.158 Mean3.83 Std. Dev.12.79 Min.−25.54 Max.73.92
  - Debt relief dummy: Obs.158 Mean0.34 Std. Dev.0.48 Dummy154
  - Storms dummy (moderate): Obs.158 Mean0.16 Std. Dev.0.37 Dummy126
  - Storms dummy (severe): Obs.158 Mean0.06 Std. Dev.0.24 Dummy110
- VAR stability: all eigenvalues have modulus inside the unit circle; VAR satisfies stability condition.
- Unit-root tests: IPS Panel Unit-root test p-values reported as 0.0000 for most series; World per capita GDP growth IPS p-value reported as 0.0092.

*Source: Excerpt from IMF working paper _wp14125 (Sections I–V, Appendix figures and tables as provided).*

### 1.    The model with ODA   .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .    19

### 1.    The model with ODA

### Introduction
- Over the past forty years, the Caribbean has suffered more than 250 natural disasters, most of which were storms.
- During this period, natural disasters killed over 12,000 people and affected over 12 million more.
- The economic impact of these disasters has been substantial with estimated damages of US$19.7 billion in 2010 constant dollars, that is, on average 1 percent of the Caribbean GDP is destroyed every year.
- The region is among the most indebted in the world, motivating the study of the effects of disasters on growth and debt.

### Research goal and scope
- Goal: determine the effects that natural disasters have on per capita GDP and on the debt to GDP ratio in the Caribbean.
- Importance: explicitly consider the role that official development assistance (ODA) and debt relief play in the aftermath of natural disasters.
- Types of disasters studied: storms and floods.
- Disaster intensity: analyze both moderate and severe disasters.
- Sample: panel covering 12 Caribbean countries over 40 years, from 1970 to 2009.
  - The countries included in the sample are: Antigua and Barbuda, The Bahamas, Barbados, Dominica, Dominican Republic, Grenada, Haiti, Jamaica, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, and Trinidad and Tobago.

### Methodology
- Econometric framework: panel vector autoregressive model with exogenous shocks (VARX).
- Estimation: panel is estimated using a bootstrap-bias correction of the fixed-effects estimator.
  - Rationale: fixed effects (LSDV) estimator captures country heterogeneity; bootstrap corrects for inconsistency of LSDV in dynamic panels when the time dimension is small.
- Advantages of panel VAR:
  - Traces the economic response in each year after a natural disaster and obtains cumulative effects after a few years.
  - Controls for other macroeconomic effects (e.g., inflation, investment, government spending) on output and debt while addressing endogeneity between variables.

### Key findings
- Output effects:
  - Both storms and floods have a negative effect on growth.
  - Severe disasters generate larger drops in output than moderate ones.
  - Excluding debt from the model overestimates the impact of disasters on output.
- Debt effects:
  - Floods increase debt.
  - Storms (in particular severe ones) do not increase debt in the full sample.
  - In the Eastern Caribbean Currency Union (ECCU) subsample, storms significantly increase debt in the short and long run.
- Role of ODA and debt relief:
  - There is weak evidence that the drop in the debt ratio after storms is partially explained by external debt relief and aid flows; storms seem to benefit from aid flows while floods do not.
  - Weak evidence that debt relief and humanitarian aid make a difference in the impact of storms on debt.
  - Debt relief appears to attenuate pressure on debt accumulation to finance reconstruction after storms, but its effect on growth is negligible.
  - Debt relief seems to have little impact on the effects that floods have on either GDP or debt.

### Advantages and limitations highlighted
- Advantages:
  - Panel VARX allows year-by-year responses and cumulative impacts.
  - Controls for macroeconomic variables and endogeneity.
- Limitations (implied by text):
  - Evidence on ODA and debt relief effects is characterized as weak.
  - Previous literature on fiscal consequences of disasters did not include Caribbean countries or control for other economic variables, potentially overestimating effects on debt.

*Source: Excerpt from the paper covering the introduction, goals, methodology, sample, and main findings (pages and figures listed in the supplied content).*

### Section III describes the data and sources, Section IV presents the econometric model and the

### _wp14125 - Section III describes the data and sources, Section IV presents the econometric model and the

### Literature review
- Summary of prior findings and methodological context:
  - Rasmussen (2004): in the Caribbean, short-run effects of disasters include an immediate contraction of output, a worsening of external and fiscal balances and an increase in transfers from abroad; long-run effects are inconclusive.
  - Cashin and Sosa (2013): climatic shocks in the ECCU lead to an immediate and significant fall in output; effects do not appear persistent.
  - Strobl (2012): using a wind field model and a destruction index similar to Emanuel (2005), finds that the average hurricane reduces output by 0.8 percentage points; notes lack of controls for inflation, investment, government spending may lead to overestimation.
  - Noy (2009): on a broad sample, developing and smaller countries are more vulnerable to disasters; better human capital, institutions, and larger governments tend to be less affected.
  - Raddatz (2009): studies different disaster types; finds ODA has only a modest and not significant reduction of disaster effects on output; initial debt level does not affect output loss from a disaster.
  - Fomby, Ikeda and Loayza (2013): use VARX, include inflation, investment, government expenditure and world output; disaggregate disaster types (droughts, floods, storms, earthquakes); find (i) droughts more negative than floods/storms/earthquakes, (ii) severe disasters often have a more negative mean response than moderate ones, (iii) timing and response vary with disaster type and sector.
- This paper’s contribution relative to prior work:
  - Follows Fomby et al.’s methodology but focuses on Caribbean countries.
  - Studies linkages between debt, growth and natural disasters.
  - Considers effects of debt relief and aid flows on debt paths after disasters.
  - Includes an additional measure of natural disaster intensity using estimated damages (see Data section).
  - Notes that Fomby et al. do not consider debt-related variables.

### Data description (Section III)
- Sample and data sources:
  - Covers 12 Caribbean countries over the 1970 to 2009 period.
  - Economic data from Penn World Tables version 7.0 (Heston, Summers, and Aten, 2011), IMF World Economic Outlook (WEO, 2011), World Bank World Development Indicators (WDI, 2011), OECD, Paris Club, and EM-DAT.
- Variables and treatment:
  - Endogenous domestic determinants (all treated as endogenous): inflation, trade openness, government consumption, investment and financial depth.
  - Exogenous external variables: world GDP growth and change in terms of trade (treated as exogenous for small open Caribbean economies).
  - Benchmark VARX includes seven endogenous variables and four exogenous variables:
    - Endogenous vector y_{i,t}:
      - Per capita GDP growth_{i,t}
      - Debt to GDP ratio growth_{i,t}
      - Investment share of GDP growth_{i,t}
      - Government share of GDP growth_{i,t}
      - Inflation rate_{i,t}
      - Trade openness growth_{i,t}
      - Financial depth growth_{i,t}
    - Exogenous vector x_{i,t}:
      - Storms_{i,t}
      - Floods_{i,t}
      - Terms of trade_{i,t}
      - World growth_{i,t}
  - Additional specifications include either ODA to GDP growth or a Debt relief dummy in x_{i,t} to control debt dynamics influenced by aid flows, debt relief and restructurings.
- Debt relief dummy construction:
  - Sources: Paris Club Agreements (for event indicator when agreement exists), World Bank Global Development Finance, OECD.Stat.
  - Dummy equals 1 when debt relief received surpasses 0.01% of GDP.
  - Also constructed an alternative dummy with threshold at least 1% of GDP; results similar but the 0.01% threshold chosen for relevance to moderate disasters.
  - Combined data sets by keeping the source that showed the largest amount in each year for each recipient country; Paris Club agreements set dummy to 1 when an agreement exists (Paris Club does not report amounts).
- Natural disaster data and intensity measures:
  - Source: EM-DAT (records type of disaster, fatalities, total affected, and when available estimated damages).
  - Two disaster categories considered: storms, and floods.
  - Primary intensity dummy measure (following Fomby et al.):
    - For each type-k natural disaster in country i, year t compute intensity^k_{i,t}:
      - intensity^k_{i,t} = { 1, if fatalities^k_{i,t} + 0.3 · total affected^k_{i,t} / population_{i,t} > 0.0001; 0, otherwise }
    - Interpretation: any natural disaster that affects more than 0.01% of the population is considered potentially disruptive.
    - The affected population is a weighted average of fatalities (weight 1) and total affected (weight 0.3); rationale: fatalities imply more severe disasters.
    - Alternative weighting tested: weight of total affected set to 1 (fatalities and total affected equal weight) for robustness.
    - Severe disaster index: threshold increased to 1% of total population to measure severe disasters.
    - Aggregation when multiple type-k disasters occur in a year: fatalities and people affected are aggregated to construct the annual intensity index.
  - Alternative intensity measure using estimated damages as percent of GDP:
    - intensity^k_{i,t} = { 1, if damages^k_{i,t} / GDP_{i,t} > 0.0001; 0, otherwise }
  - Notes on EM-DAT inclusion criteria: a disaster is included if at least one of the following is met: i) at least 10 people killed, ii) at least 100 people affected, or iii) a state of emergency declared or a call for international assistance made.
- Decision rule for modeling disasters:
  - Uses dummy for a "disastrous year" (presence of intensity = 1) rather than counting additional disasters, to better match annual debt decision-making and reconstruction responses.
  - Author reports estimation results using Fomby et al.’s disasters measure are similar (results not included but available on request).

### Econometric model (Section IV)
- Baseline model: fixed effects unbalanced panel VAR with exogenous variables (VARX), following Fomby et al.
  - Model equation (3):
    - y_{i,t} = α_i + Φ_1 y_{i,t−1} + Φ_2 y_{i,t−2} + Θ_0 x_{i,t} + Θ_1 x_{i,t−1} + Θ_2 x_{i,t−2} + ε_{i,t}
    - ε_{i,t} is a vector of errors; countries indexed by i = 1,2,...,N; time index for each country i is t = 1,...,T_i.
- Fixed effects setup:
  - α_i captures unobserved time-invariant heterogeneity.
  - Stack observations over time and cross-sections to write y = D α′ + Z δ′ + ε where Z = (y_{−1} y_{−2} x x_{−1} x_{−2}).
  - Least-squares dummy variable (LSDV) estimator for δ:
    - δ̂ = (Z′ A Z)^{-1} Z′ A y where A is block-diagonal with A_i = I_{T_i} − (1/T_i) i_{T_i} i_{T_i}′.
  - Fixed-effects coefficients recovered as α̂ = (D′ D)^{-1} D′ (y − Z δ̂).
- Mean response (generalized impulse response):
  - After controlling for fixed effects, multiplier form:
    - y_{i,t} = Φ(L)^{-1} Θ(L) x_{i,t} + Φ(L)^{-1} ε_{i,t}
  - Mean response of y to x is Ψ(L) = Φ(L)^{-1} Θ(L).
  - Interpretation: generalized impulse response function (Pesaran and Shin, 1998) measuring direct impact of disasters on each endogenous variable at time zero (lower bound since secondary endogenous feedbacks are not attributed at time zero).
- Inference for impulse responses:
  - Standard errors computed via Monte Carlo method (Hamilton, 1994):
    - Draw π^(1) ~ N(π̂, (1/T)(Ω̂ ⊗ Q̂^{-1})) where ψ_s ≡ vec(Ψ_s), π̂ ≡ vec(δ̂), Q̂^{-1} = [(1/T)(Z′ Z)]^{-1}.
    - Compute ψ_s(π^(1)) and repeat 30,000 times to build simulation-based confidence intervals (e.g., 90% CI from simulated distribution).
- Stability condition:
  - VAR stability requires all eigenvalues of the companion matrix F have modulus less or equal than 1 with |λ| < 1. Table 5 (in the source) shows the VAR is stable.
- Note on LSDV bias and correction:
  - Because model is dynamic and T_i is small and fixed, the LSDV estimator is inconsistent (Nickell, 1981); bias decreases as T_i grows.
  - Bootstrap bias-corrected (BSBC) estimator implemented following Pesaran and Zhao (1999), Everaert and Pozzi (2007), and Fomby et al.
  - BSBC algorithm summary:
    1. Estimate δ̂ using LSDV, obtain α̂, and calculate residuals ε̂ = A y − A Z δ̂.
    2. Set number of bootstrap samples B = 1,000. For j = 1,...,B:
       - (a) Obtain a bootstrap sample ε̂^{b(j)} of the rescaled residuals ε̂_r by choosing T integers between 1 and T at random with equal probability. Rescaled residuals:
         - ε̂_{r,i,t} = √(T_i/(T_i − 1)) ( ε̂_{i,t}/√m_{i,t} − (1/T_i) Σ_{k=1}^{T_i} ε̂_{i,k}/√m_{i,k} ) for t = 1,...,T_i and i = 1,...,N;
         - where m_{i,t} is the it-th diagonal element of the projection matrix M = (I_T − A Z (Z′ A Z)^{-1} Z′) A.
       - (b) Calculate bootstrap sample y^{b(j)} = D α̂′ + Z^{b(j)} δ̂′ + ε̂^{b(j)}, where Z^{b(j)} = (y^{b(j)}_{−1} y^{b(j)}_{−2} x x_{−1} x_{−2}); initialize y^{b(j)}_{i,−1} = y_{i,−1} and y^{b(j)}_{i,0} = y_{i,0}.
       - (c) Obtain LSDV estimator δ̂^{b(j)} = (Z^{b(j)′} A Z^{b(j)})^{-1} Z^{b(j)′} A y^{b(j)} and mean responses Ψ̂^{b(j)}(L) = [Φ̂^{b(j)}(L)]^{-1} Θ̂^{b(j)}(L).

*Source: _wp14125 - Section III describes the data and sources, Section IV presents the econometric model and the (IMF working paper content provided).*

### 3.  Then calculate the bootstrap bias-corrected (BSBC) mean response

### 3.  Then calculate the bootstrap bias-corrected (BSBC) mean response

### BSBC estimator and standard errors
- BSBC mean response for the s-lag:
  - ̃Ψs = 2 ˆΨs − 1 B B ∑j=1 ˆΨb(j) s, for s = 1,2,...
  - ˆΨs is the LSDV mean response of Ψs.
- Procedure to calculate standard errors of BSBC impulse-responses:
  - Generate B bootstrapped samples for each ˆδb(j) and calculate the BSBC mean response ̃Ψ(j) for each sample using the BSBC algorithm to obtain {̃ψ(1), ..., ̃ψ(B)}.
  - If 90% of the simulations for the first element of ψs lie between (̃ψs1, ̃ψs1) this interval serves as the 90% confidence interval for the first element of ̃ψs.
- Implementation note:
  - The model is estimated in Matlab building on the code developed by Fomby et al.

### Diagnostic tests: stationarity, lag selection, exogeneity
- Stationarity tests:
  - Im, Pesaran, Shin (IPS) panel unit-root tests used for unbalanced panels.
  - Results (column 1 of Table 6): the null that all panels have a unit-root is rejected in all cases.
  - Individual ADF and PP tests show the majority of countries do not exhibit signs of unit-roots.
  - Columns 2 and 3 of Table 6 present the percentage of countries that reject the presence of a unit-root for each variable.
  - Conclusion: variables, in log differences, are appropriate for a VARX model.
- Lag structure selection:
  - Use AIC and SBIC information criteria and likelihood ratio tests; test up to three lags for endogenous and exogenous variables (p and q).
  - Information criteria point to p = q = 1.
  - Likelihood ratio tests suggest p = q = 3.
  - Compromise choice: p = q = 2 for all estimations.
  - Rationale: likelihood ratio tests indicate one-lag VARX does not capture dynamics well; information criteria indicate information loss with 3 lags is higher.
- Granger causality tests for ODA to GDP growth and debt relief dummy:
  - Tests (Table 8) show both variables are not Granger caused by endogenous variables in the model → treated as exogenous and included in xi,t vector.
  - Debt relief exogeneity is consistent with external shocks like the HIPC initiative changes in 1999 and donor behavior (examples cited: St. Vincent and the Grenadines hurricane Tomas October 2010 and floods April 2011).
  - ODA exogeneity result is surprising given ODA includes humanitarian aid, debt relief, and development projects.

### V. Results — overview and methodology
- Focus:
  - Dynamic effects and adjustment path after natural disasters (storms and floods) on GDP growth and debt.
  - Use mean impulse response functions and mean cumulative response of GDP growth and debt ratio growth to each disaster type.
  - Present mean and cumulative mean responses for years 0 through 8 (0 = year of disaster) with 10% tails confidence intervals.
- Figures:
  - Figures 1 to 10 present mean and cumulative responses; first column = storms, second column = floods; upper half = GDP, bottom half = debt ratio.
  - Results correspond to the BSBC estimator.
- Confidence intervals:
  - Constructed using the procedure explained in the previous section.

### A. The benchmark model — effects on output and debt
- Immediate effects on output growth:
  - Both storms and floods have an immediate negative effect, but none appear statistically significant.
  - Moderate disasters:
    - Moderate storms: on impact GDP growth falls by more than half a percentage point.
    - Moderate floods: on impact GDP growth falls by 0.1 percentage points.
  - Severe disasters:
    - Severe floods: immediate effect close to 3 percentage points drop in GDP growth.
    - Severe storms: immediate effect above 1 percentage point drop in GDP growth.
- Recovery path:
  - Small rebound in economic activity:
    - Year 1 after storms.
    - Year 2 after floods (year 3 after severe floods).
  - Rebound is short lived; relapse to negative growth follows after about a year.
  - After the fifth or sixth year effects are small and tend to disappear.
- Cumulative effects on output:
  - Moderate storms: economy remains about 0.5 percent below the level without the disaster (no full recovery).
  - Moderate floods: cumulative effect slightly positive on output level.
  - For both storms and floods cumulative effects are not statistically significant.
  - Severe storms: recovery in first few years yields economy slightly above non-disaster output level.
  - Severe floods: only a small rebound and cumulative effect on output is negative.
- Interpretation of different government spending and reconstruction dynamics:
  - Moderate storms: limited infrastructure damage, larger effects on livelihoods → increased social assistance with small growth effect; short recovery (~1 year).
  - Severe storms: larger infrastructure damage → big reconstruction investments that support recovery; longer recovery (~5 years).
  - Moderate floods: potential lasting positive effect via higher land productivity (Fomby et al.), but in Caribbean sample moderate floods show initial negative effect (not statistically significant), possibly due to small agricultural sectors.
- Role of public debt in estimation:
  - Replication excluding debt (Fomby et al. model) overestimates negative effect of disasters on GDP.
  - Excluding debt biases disaster estimates on growth because part of negative effect is captured by debt.
  - Including public debt is crucial to understand dynamic effects of natural disasters.
- Effects on debt:
  - Moderate and severe floods increase debt ratio growth rate.
    - Severe floods: significant increase in growth of debt ratio of about 16 percentage points.
    - Increase in debt after severe floods not reflected in pickup of economic activity until third year of recovery.
  - Severe storms:
    - Effect on debt is negative and significant; debt ratio seems to decline after a storm.
    - Decline in growth rate of debt ratio lasts:
      - Two periods after a moderate storm.
      - Three periods after a severe storm.
    - Severe storm initial decline in debt growth rate is significant and close to 8 percentage points.
  - Explanation for decline in debt after severe storms explored in next subsection.

### B. The role of debt relief and ODA
- Method:
  - Use two alternative model specifications: one with ODA as exogenous, another with debt relief dummy as exogenous.
  - Debt relief dummy comprises debt forgiveness, debt restructuring and humanitarian aid flows > 0.01% of recipient’s country GDP.
- 1. Model with ODA:
  - Results (Figures 3 and 4) show no material difference compared to benchmark (Figures 1 and 2).
  - Consistent with Raddatz (2009): ODA has a modest and not significant reduction of disaster effects.
  - Explanation: ODA includes long-term projects not influenced by short-term disasters in addition to humanitarian aid and debt relief.
- 2. Model with debt relief:
  - Results (Figures 5 and 6) differ from benchmark but are not statistically significant → treated as weak evidence.
  - Growth:
    - Very similar to cases without debt relief included.
  - Debt:
    - After controlling for debt relief, first 3 years after a moderate storm show an initial increase in debt (not significant).
    - Cumulative effect shows increase in debt ratio level of about 3 percent when debt relief included; contrasts with negligible effect when debt relief excluded.
    - Interpretation: debt relief may help finance reconstruction by ameliorating demand to increase public debt, but evidence is not conclusive.
    - After a severe storm, controlling for debt relief still shows a significant decline in debt ratio growth, but smaller (7.6 percentage points) than without debt relief (7.9 percentage points).
    - Flood results remain unaltered: severe flood initial mean response of debt to GDP ratio growth is an increase close to 15 percentage points.
  - Note: exclusion of debt dummy biases estimates of disasters on debt to GDP ratio by capturing part of negative effect of debt relief.

### C. Storms in the Eastern Caribbean Currency Union (ECCU)
- Sample and model specifics:
  - ECCU subsample: 6 countries.
  - Only storms estimated (few floods: three moderate, zero severe).
  - Model estimated with p = q = 1 due to smaller sample size.
  - Results for moderate and severe storms presented in Figures 7 to 10.
- Moderate storms in ECCU:
  - Immediate effect: decline in GDP growth of about 0.7 percentage points (not statistically significant).
  - Unlike broader Caribbean, ECCU does not show strong recovery after moderate storms; economies decline further next year.
  - Cumulative impact: borderline significant lower GDP level of about 2 percent two years after a storm.
  - Debt effect: during the year of the storm the debt to GDP ratio grows faster by almost 5 percentage points.
    - Cumulative effect: debt to GDP level more than 5 percent higher after 8 years.
  - Interpretation: ECCU countries are smaller and more vulnerable → larger portion affected, limited response capacity without increasing debt; increased debt may not support recovery and may be used for social assistance with little growth impact.
- Severe storms in ECCU:
  - Immediate effect: decline in growth close to 2 percentage points (borderline significant) with negative cumulative effect due to very small recovery.
  - Debt: severe storms increase debt ratio growth by almost 2 percentage points on impact; cumulative increase in debt of 5 percent by year 8 (results not statistically significant).
  - Contrast with broader Caribbean:
    - Broader Caribbean: severe storms have positive cumulative effect on output and decline in debt.
    - ECCU: debt increases, reconstruction activity lower, cumulative GDP effect is negative.

*Source: _wp14125 - 3.  Then calculate the bootstrap bias-corrected (BSBC) mean response*

### 1.  The role of debt relief

### 1.  The role of debt relief

### Role and effects of debt relief
- The positive effects of debt relief after a natural disaster are similar in the case of the ECCU (Figures 9 and 10).
- The effects on growth for both moderate and severe storms appear to be small.
- After controlling for the effect of debt relief the impact on debt is larger both in the year of the event and over the long run for moderate and severe storms.
- There is weak evidence that debt relief helps financing the costs associated with a natural disaster, and reducing the impact on debt.
- Conclusion: debt relief seems to play a role in easing the negative consequences of natural disasters in the Caribbean, and in helping the countries in the reconstruction phase.

### Robustness of results
- All results presented were obtained from the BSBC estimator; results were also estimated using the LSDV estimator as a robustness check.
- The LSDV results turned out to be very similar to the BSBC ones, suggesting that given the average time sample of 31 years in the data, the LSDV estimator is a good approximation (the LSDV estimator bias is small).
- The LSDV results are not included in the paper but are available upon request.
- Appendix robustness checks:
  - Figures A3 and A4 use total affected as an alternative measure of disaster intensity.
  - Figures A5 and A6 use the estimated damages in U.S. dollars to construct the disaster intensity dummy variable.
- Robustness finding: the qualitative response to a disaster at year 0 is robust to the different measures of disaster intensity; however, the cumulative effects seem to depend on the measure used.

### Main conclusions (summary of empirical findings)
- Methodology: panel VAR with exogenous variables; three models estimated:
  - benchmark model analyzing moderate and severe storms and floods on debt and growth;
  - two alternative models analyzing the effects of debt relief and ODA after disasters.
- General findings:
  - Both storms and floods have a negative effect on growth, with severe disasters generating larger drops in output.
  - Including debt in the model is important because its exclusion overestimates the impact of disasters on output (when debt is excluded the coefficients for disasters capture part of the negative effect of debt on growth).
  - The increase in the debt ratio after a disaster is driven both by lower GDP and by an increase in debt levels to finance recovery and reconstruction.
  - Results indicate that floods increase debt while storms (in particular severe ones) do not—partly explained by the role of external debt relief and aid flows in recovery where storms seem to benefit from aid flows while floods do not; evidence is weak and far from conclusive.
  - Neither floods nor storms (separately or combined) are good predictors of a country receiving debt relief—interpret results carefully.
- ECCU-specific findings:
  - In the ECCU, moderate and severe storms increase debt in the short run and in the long run; only the effect of moderate storms is statistically significant.
  - The effect of storms on growth in the ECCU is negative; with severe storms initially having a larger negative effect.
  - Higher debt does not seem to support stronger economic activity during the recovery phase in either moderate or severe disasters.
  - Hypothesis: moderate disasters likely require more expenditure on social assistance, while severe ones require a larger component of infrastructure investment that is more supportive of economic rebound; nevertheless, replacing damaged capital goods has an opportunity cost versus investing in new capital, potentially weakening recovery when it involves increasing debt. More research is needed.

### Role of aid and targeted flows
- General aid flows (ODA) do not have much of an effect in the aftermath of natural disasters.
- “Targeted aid flows” (debt relief, debt restructuring, and humanitarian aid, captured in the debt relief dummy) after a storm seem to have a role in lessening the negative effects on debt.
- There is weak evidence that debt relief helps reduce the pressure on debt accumulation to finance reconstruction activity after storms.
- Evidence suggests debt relief flows are exogenously determined (they do not respond to the economic conditions of the receiving country), and the relief response seems to be uneven among disasters.

### Policy implications and recommendations
- Disaster-prone countries should:
  - expect debt to rise in the aftermath of disasters (especially if they don’t receive foreign aid),
  - attempt to keep sustainable debt levels, and create enough fiscal space to be able to finance their own recovery.
- Other measures to pursue:
  - establishing disaster funds,
  - increasing insurance coverage for the public and private sector,
  - building resilience by integrating climate and disaster risk into policy frameworks at both the macro and micro levels.
- Note: For an in depth analysis of options available to governments in the aftermath of disasters see Cashin and Dyczewski (2006) (referenced in the source); for comprehensive analyses on building resilience see World Bank (2013) and Laframboise and Loko (2012) (referenced in the source).

### Selected key statistics and data notes (as reported)
- Average time sample in the data: 31 years.
- Table 2 (as presented): Antigua and Barbuda40305; The Bahamas50000; Barbados41006; Dominica603013; Dominican Republic872113; Grenada401013; Haiti782016; Jamaica872223; St. Kitts and Nevis40205; St. Lucia30006; St. Vincent and the Grenadines531012; Trinidad and Tobago30103; Total6126173115
- Table 3 (selected rows as presented):
  - Per capita GDP growth: Obs.359 Mean2.39 Std. Dev.5.37 Min.−19.82 Max.18.77
  - Debt ratio growth: Obs.359 Mean3.01 Std. Dev.16.08 Min.−100.48 Max.77.77
  - Debt relief dummy: Obs.359 Mean0.32 Std. Dev.0.47 Dummy1115
  - Storms dummy (moderate): Obs.359 Mean0.17 Std. Dev.0.38 Dummy161
  - Floods dummy (moderate): Obs.359 Mean0.07 Std. Dev.0.26 Dummy126
  - Storms dummy (severe): Obs.359 Mean0.05 Std. Dev.0.21 Dummy117
  - Floods dummy (severe): Obs.359 Mean0.01 Std. Dev.0.09 Dummy13
- Table 4 (ECCU selected rows as presented):
  - Per capita GDP growth: Obs.158 Mean3.43 Std. Dev.5.05 Min.−13.26 Max.15.88
  - Debt ratio growth: Obs.158 Mean3.83 Std. Dev.12.79 Min.−25.54 Max.73.92
  - Debt relief dummy: Obs.158 Mean0.34 Std. Dev.0.48 Dummy154
  - Storms dummy (moderate): Obs.158 Mean0.16 Std. Dev.0.37 Dummy126
  - Storms dummy (severe): Obs.158 Mean0.06 Std. Dev.0.24 Dummy110
- VAR stability: all modulus of the eigenvalues lie inside the unit circle, therefore the VAR satisfies the stability condition.
- Unit-root tests: IPS Panel Unit-root test p-values reported as 0.0000 for most series; World per capita GDP growth IPS p-value reported as 0.0092.
- Lag structure selection and Granger causality tests summarized in the paper indicate exogeneity of ODA and the debt relief dummy with respect to the endogenous variables (see tables and notes in the source).

*Source: _wp14125 - 1.  The role of debt relief (PDF chapter) — content as provided*

### REFERENCES

### _wp14125 - REFERENCES

### References
- Abbas, Ali; Belhocine, Nazim; ElGanainy, Asmaa; and Horton, Mark, 2010, “A Historical Public Debt Database,” IMF Working Paper 10/245 (Washington: International Monetary Fund).
- Cashin, Paul; and Dyczewski, Pawel, 2006, “Government Responses to Natural Disasters in the Caribbean,” inThe Caribbean: From Vulnerability to Sustained Growth, edited by R. Sahay, D. Robinson and P. Cashin (Washington: International Monetary Fund).
- Cashin, Paul; and Sosa, Sebastian, 2013, “Macroeconomic Fluctuations in the Eastern Caribbean: the Role of Climatic and External Shocks.”Journal of International Trade &Economic Development, Vol. 22, pp. 729-748.
- Eisensee, Thomas; and Strömberg, David, 2007, “News droughts, news floods, and U.S. disaster relief.”The Quarterly Journal of Economics, Vol. 122, pp. 693-728.
- EM-DAT: The OFDA/CRED International Disaster Database. Université Catholique de Louvain –Brussels– Belgium. Available in: www.emdat.be
- Emanuel, Kerry, 2005, “Increasing destructiveness of tropical cyclones over the past 30 years.”Nature, Vol. 436, pp. 686-688.
- Everaert, Gerdie; and Pozzi, Lorenzo, 2007, “Bootstrap-based bias correction for dynamic panels,”Journal of Economic Dynamics and Control, Vol. 31 (4), pp. 1160-1184.
- Fomby, Thomas; Ikeda, Yuki; and Loayza, Norman, 2013, “The Growth Aftermath of Natural Disasters,”Journal of Applied Econometrics, Vol. 28 (3), pp. 412-434. Data and codes available in: http://qed.econ.queensu.ca/jae/datasets/fomby001/
- Global Development Finance, December 2011. World Bank. Available in: http://databank.worldbank.org/ddp/home.do?Step=12&id=4&CNO=2
- Green, William, 2008,Econometric Analysis(New Jersey: Pearson Prentice Hall, 6th ed.).
- Hamilton, James, 1994,Times Series Analysis(Princeton: Princeton University Press).
- Heston, Alan; Summers, Robert; and Aten, Bettina, 2011, Penn World Tables v.7.0.Center for International Comparisons of Production, Income and Pricesat the University of Pennsylvania.
- IMF, 2011a, “St. Vincent and the Grenadines: Request for Disbursement Under the Rapid Credit Facility” IMF Country Report No. 11/349 (Washington: International Monetary Fund).
- IMF, 2011b, “St. Vincent and the Grenadines: Request for Disbursement Under the Rapid Credit Facility” IMF Country Report No. 11/344 (Washington: International Monetary Fund).
- Laframboise, Nicole; and Loko, Boileau, 2012, “Natural Disasters: Mitigating Impact, Managing Risks,” IMF Working Paper 12/245 (Washington: International Monetary Fund).
- Nickell, Stephen, 1981, “Biases in Dynamic Models with Fixed Effects.”Econometrica, Vol. 49, pp. 1417-1426.
- Noy, Ilan, 2009, “The macroeconomic consequences of disasters.”Journal of Development Economics, Vol. 88, pp. 221-231.
- Noy, Ilan; and Nualsri, Aekkanush, 2011, “Fiscal Storms: Public spending and revenues in the aftermath of natural disasters.”Environment and Development Economics, Vol. 16, pp. 113-128.
- OECD.Stat, March 2012. Organization for Economic Co-operation and Development. Available in: http://www.oecd-ilibrary.org/statistics
- Paris Club Agreements, March 2012. Paris Club. Available in: http://www.clubdeparis.org/en/
- Pesaran, Hashem; and Shin, Yongcheol, 1998, “Generalized impulse response analysis in linear multivariate models.”Economics Letters, Vol. 58, pp. 17-29.
- Pesaran, Hashem; and Zhao, Zhongyun, 1999, “Bias reduction in estimating long-run relationships from dynamic heterogeneous panels.” InAnalysis of Panels and Limited Dependent Variable Models, Hsiao C, Pesaran MH, Lahiri K, Lee LF (eds). Cambridge University Press: Cambridge, UK; pp. 297-322.
- Raddatz, Caludio, 2009, “The Wrath of God: Macroeconomic Costs of Natural Disasters,” Policy Research Working Paper 5039 (Washington: World Bank).
- Rasmussen, Tobias, 2004, “Macroeconomic Implications of Natural Disasters in the Caribbean,” IMF Working Paper 04/224 (Washington: International Monetary Fund).
- Rasmussen, Tobias, 2006, “Natural Disasters and Their Macroeconomic Implications,” inThe Caribbean: From Vulnerability to Sustained Growth, edited by R. Sahay, D. Robinson and P. Cashin (Washington: International Monetary Fund).
- Sims, Christopher, 1980, “Macroeconomics and Reality.”Econometrica, Vol. 48, pp.1-48.
- Skidmore, Mark; and Toya, Hideki, 2002, “Do natural disasters promote long-run growth?” Economic Inquiry, Vol. 40, pp. 664-687.
- Strobl, Eric, 2012, “The economic growth impact of natural disasters in developing countries: Evidence from hurricane strikes in the Central American and Caribbean regions.”Journal of Development Economics, Vol. 97, pp. 130-141.
- Toya, Hideki; and Skidmore, Mark, 2007, “Economic development and the impacts of natural disasters.”Economics Letters, Vol. 94, pp. 20-25.
- World Bank, 2013, “Building Resilience: Integrating Climate and Disaster Risk into Development,” Lessons from World Bank Group Experience (Washington: World Bank).
- World Development Indicators, December 2011. World Bank. Available in: http://databank.worldbank.org/ddp/home.do?Step=12&id=4&CNO=2
- World Economic Outlook Database, September 2011. International Monetary Fund. Available in: http://www.imf.org/external/pubs/ft/weo/2011/02/weodata/index.aspx

### Appendix A. Additional Figures
- Figure A1. Mean Responses of GDP to Moderate Disasters
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4

- Figure A2. Mean Responses of GDP to Severe Disasters
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −8 −6 −4 −2 0 2 4
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −8 −6 −4 −2 0 2 4
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −20 −15 −10 −5 0 5 10
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −20 −15 −10 −5 0 5 10

- Figure A3. Mean Responses of GDP and Debt to Moderate Disasters (Using Total Affected as an Alternative Measure of Intensity)
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4
  - Mean response of debt ratio growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10
  - Mean response of debt ratio growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10
  - Cumulative mean response of debt ratio to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10 15
  - Cumulative mean response of debt ratio to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10 15

- Figure A4. Mean Responses of GDP and Debt to Moderate Disasters Including the Debt Relief Dummy as an Exogenous Variable (Using Total Affected as an Alternative Measure of Intensity)
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −3 −2 −1 0 1 2 3
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −3 −2 −1 0 1 2 3 4
  - Mean response of debt ratio growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10
  - Mean response of debt ratio growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10
  - Cumulative mean response of debt ratio to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10 15
  - Cumulative mean response of debt ratio to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −10 −5 0 5 10 15

- Figure A5. Mean Responses of GDP and Debt to Moderate Disasters (Using Damages as an Alternative Measure of Intensity)
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −2 0 2 4 6
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −2 0 2 4 6
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −5 0 5 10
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −5 0 5 10
  - Mean response of debt ratio growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −25 −20 −15 −10 −5 0 5 10 15
  - Mean response of debt ratio growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −25 −20 −15 −10 −5 0 5 10 15
  - Cumulative mean response of debt ratio to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −40 −30 −20 −10 0 10 20
  - Cumulative mean response of debt ratio to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −40 −30 −20 −10 0 10 20

- Figure A6. Mean Responses of GDP and Debt to Moderate Disasters Including the Debt Relief Dummy as an Exogenous Variable (Using Damages as an Alternative Measure of Intensity)
  - Mean response of GDP growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −2 0 2 4 6
  - Mean response of GDP growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −4 −2 0 2 4 6
  - Cumulative mean response of GDP to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −5 0 5 10
  - Cumulative mean response of GDP to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −5 0 5 10
  - Mean response of debt ratio growth to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −25 −20 −15 −10 −5 0 5 10 15
  - Mean response of debt ratio growth to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −25 −20 −15 −10 −5 0 5 10 15
  - Cumulative mean response of debt ratio to storm shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −40 −30 −20 −10 0 10 20
  - Cumulative mean response of debt ratio to flood shock (Time (years)): 0 1 2 3 4 5 6 7 8 with axis values −40 −30 −20 −10 0 10 20

*Source: _wp14125 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14125.pdf_