## How Do International Financial Flows to Developing Countries Respond to Natural Disasters? ## Source details **Canonical URL:** [How Do International Financial Flows to Developing Countries Respond to Natural Disasters?](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10166.pdf) ## Other formats - [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10166.pdf.md) - [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10166.pdf.json) --- ### Abstract and Key Conclusions - Uses multivariate dynamic panel analysis (panel vector autoregression) to examine the response of international financial flows to natural disasters. - Sample: a large sample of developing countries covering the period from 1970 to 2005. - Main empirical findings: - Remittance inflows increase significantly in response to shocks to both climatic and geological disasters. - Responses of aid flows to natural disaster shocks in general tend not to be statistically significant. - International assistance to low income countries increases following geological disaster shocks. - Other private capital flows (bank lending and equity) typically do not attenuate the effects of disasters and in some specifications even amplify the negative economic effects of these events. - Policy implication: - Countries should take their vulnerability to natural disasters into account when considering the costs and benefits of the liberalization of private capital flows. ### Motivation, Objectives, and Theoretical Links - Motivation: - Climate change likely leads to increased variability and more frequent and intense extreme weather events; developing countries are more vulnerable. - Prior macro evidence: a one standard deviation increase in direct damages can reduce output growth by about 9 percent in a developing country (Noy, 2009). - Objectives: - Assess responses of development aid, migrant remittances, equity flows and bank lending flows to natural disaster shocks. - Examine whether capital flows exacerbate or attenuate economic effects of natural disasters. - Inform policy on financing reconstruction and roles of foreign aid versus private capital. - Theoretical transmission channels summarized: - Neo-classical: destruction of capital raises marginal product of capital and may attract inflows. - Countervailing forces: destruction of complementary inputs (infrastructure, human capital) can lower returns and cause outflows. - Market frictions and political instability can impede inflows; governments may borrow abroad (concessional aid, bank loans, bonds) to smooth. ### Empirical Framework and Data - Empirical approach: - Panel VAR specification: A0 z_it = sum_{j=1}^q A_j z_{i,t-j} + theta_i + epsilon_it. - z_it comprises endogenous variables y_it and weakly exogenous variables x_it. - Natural disaster shocks treated as weakly exogenous; structural identification requires assumptions on contemporaneous A0. - Estimation: GMM-panel VAR (PVAR) with lagged variables as instruments; forward mean differencing (Helmert) used in PVAR. - Data coverage and construction: - Sample period: 1970-2005. - Capital flows: ODA, bank and equity flows from World Bank’s Global Development Finance; remittances from World Bank’s Development Prospects Group. - Natural disaster data: OFDA/CRED EM-DAT. - Capital flows expressed in real dollar terms (deflated by US CPI index). - Key variables: log real interest rate differential, ∆RER = first difference of log real trade-weighted effective exchange rate, Ydiff = log income differential vs United States (PWT). - Natural disaster classification: Climatic (floods, droughts, extreme temperatures, hurricanes); Geological (earthquakes, landslides, volcanic eruptions, tidal waves); Human (famines, epidemics). - EM-DAT inclusion criteria: at least one of (a) 10 or more people killed; (b) 100 or more people affected; (c) a state of emergency declared; (d) a call for international assistance. ### Stationarity and Estimation Choices - Panel dimensions: Number of panels = 78; Number of periods = 36. - Unit-root test results (preserve exact reported values): - Levin-Lin-Chu (Adjusted t*): Remittances: -13.31, p-value 0.00; ODA: -10.08, p-value 0.00; Bank: -18.28, p-value 0.00; Equity: -3.35, p-value 0.00; Income Differential: -5.17, p-value 0.00; Δ Real Exchange Rate: -22.01, p-value 0.00; Interest Differential: -13.23, p-value 0.00. - Breitung (lambda): Remittances: -2.93, p-value 0.00; ODA: -4.55, p-value 0.00; Bank: -18.60, p-value 0.00; Equity: -6.93, p-value 0.00; Income Differential: 2.66, p-value 0.99; Δ Real Exchange Rate: -14.62, p-value 0.00; Interest Differential: -8.41, p-value 0.00. - Im-Pesaran-Shin (Z-t-bar): Income Differential: -5.27, p-value 0.00; Δ Real Exchange Rate: -18.79, p-value 0.00; Interest Differential: -14.93, p-value 0.00. (Remittances/ODA/Bank/Equity listed as n.a. in IPS table.) - Interpretation: - Tests strongly suggest remittances, net development aid, net bank loans, net equity flows, interest differential and first difference of the real effective exchange rate are stationary. - Income differential tests ambiguous (LLC and IPS indicate stationarity; Breitung suggests unit roots). - Overall conclusion: estimation of Panel VAR models in levels is appropriate. ### Main Empirical Findings — Summary by Flow Type - Remittances: - Tend to increase following climatic and geological disasters; do not respond to human disasters. - Example elasticities (from semi-elasticities): - Remittances typically increase by 0.1 percent in the same time period following an increase of one standard deviation in climatic disasters (evaluated at the mean). - For a country at the 95th percentile of climatic disaster incidence, contemporaneous elasticity to a standard deviation increase ≈ 0.2 percent. - For geological disasters, a country in the 95th percentile would experience an elasticity ≈ 0.1 percent for a one standard deviation increase. - Fixed-effects regressions show strong persistence in remittances: lag coefficients around 0.817*** to 0.895*** across specifications. - PVAR results: remittances increase on impact to climatic shocks (persist ~1 year) and to geological shocks (positive and marginally significant up to two years). FEVD: climatic shocks explain about 16 percent of remittance variance at 10-year horizon (full sample); geological shocks account for 34 percent (full sample). In LICs, climatic shocks account for over 50 percent of FEVD for remittances. - Official Development Assistance (ODA): - In most specifications, aid does not respond statistically significantly to disasters. - Where significant: aid increases more slowly; example — aid increases by around 0.3 percent two years after a one standard deviation increase in geological disasters for a country in the 95th percentile. - PVAR: no statistically significant response to climatic shocks in full sample; aid increases on impact following geological disasters in LICs. FEVD: geological events account for 11 percent of variance of aid flows to LICs. - Bank lending flows: - Bank lending flows respond negatively and contemporaneously to climatic disasters and to human disasters; effects are typically short-lived. - Reported magnitudes: - Net bank lending flows typically decrease by around 0.8 percent within the same time period following a one standard deviation increase in climatic disasters. - For human disasters, bank flows decrease by about 0.1 percent following an increase of one unit in incidence. - PVAR: negative and significant response on impact to climatic and geological disasters; often becomes indistinguishable from zero in subsequent periods. FEVD: geological shocks account for 23 percent of variation in bank flows at 10-period horizon; climatic shocks account for 2 percent. - For LICs, climatic events account for 13 percent of FEVD for bank flows (t=10) while geological disasters account for 5 percent (text note). - Net equity flows: - Increase following climatic and geological disasters typically with a one-year lag; responses to human disasters can appear with a two-year lag. - Magnitudes: - Net equity flows increase by around 0.5 percent one year after a one standard deviation increase in climatic disasters. - For a country in the 95th percentile of climatic disasters, the increase is 1.3 percent one year after a one standard deviation increase. - Net equity flows increase by 0.1 percent one year after a one standard deviation increase in geological disasters. - Equity flows increase by about 0.1 percent two years after a one standard deviation increase in human disasters. - Equity responses tend to be short-lived; for LICs equity flows are generally not a significant source of disaster recovery finance. ### PVAR and Forecast Error Variance Decomposition (selected exact numbers) - FEVD — Bank flows (variance explained at t=10): - Climatic Events: 0.02 n.a. 0.13 n.a. - Income Differential: 0.02 0.03 0.01 0.01 - Real Exchange Rate: 0.00 0.00 0.00 0.00 - Interest Differential: 0.01 0.01 0.00 0.00 - Bank: 0.95 0.73 0.86 0.94 - Geological Events: n.a. 0.23 n.a. 0.05 - FEVD — Equity flows (variance explained at t=10): - Climatic Events: 0.29 n.a. 0.00 n.a. - Income Differential: 0.03 0.01 0.01 0.01 - Real Exchange Rate: 0.02 0.01 0.00 0.00 - Interest Differential: 0.00 0.00 0.00 0.00 - Equity: 0.66 0.93 0.99 0.99 - Geological Events: n.a. 0.05 n.a. 0.00 - Additional FEVD findings: - Climatic disasters account for 13 percent of the FEVD of bank flows for LICs (text). - Climatic shocks account for about 29 percent of the forecast error in equity flows (full sample). - Natural disasters can account for up to 53 percent of the forecast error variance of remittance inflows to low-income countries (conclusions). ### Robustness, Model Fit, and Caveats - Robustness notes: - Inclusion of deterministic trends and ∆RER can affect significance and reduce sample size; some disaster effects lose statistical significance in these specifications. - Bank and equity flow models show relatively poorer goodness-of-fit, indicating omitted-variable concerns. - Many developing countries in sample lack access to international equity markets; equity-flow results have limited generalizability for LICs. - Methodological caveats: - GMM estimator may be less efficient and biased if instruments are weak. - PVAR GMM assumes homogeneous dynamics across countries; estimates biased if dynamics differ. - Natural disaster measures could be refined to capture variation in economic impact magnitude. - Differentiation between types of aid (budget support vs project) could change conclusions. - Suggested future work: test results with alternative estimators (e.g., mean-group) and refine disaster measures. ### Policy Implications and Recommendations - Remittances: - Foster remittance flows for disaster-vulnerable countries (for example, by reducing remittance transaction costs) because remittances act as a compensatory flow after climatic and geological shocks. - Foreign aid: - Policymakers should not generally rely on substantial increases in foreign assistance to finance reconstruction or consumption smoothing needs, except in specific circumstances (aid shows limited response in the heterogeneous full sample but increases to geological shocks in LICs). - International community should strengthen capacity to scale-up aid after disasters; consider recipient absorptive capacity when scaling up aid. - Capital account and private flows: - Consider vulnerability to natural disasters when weighing costs and benefits of capital account liberalization. - Benefits of liberalization (facilitating financing for disaster recovery) typically do not appear to have materialized for many developing countries; there is evidence of risks of private capital outflows after disasters. - General: - Policy design should account for heterogeneity across disaster types and across income groups; remittances provide immediate relief, aid is more relevant for geological disasters in LICs, and private bank flows can contract on impact. ### Data Summary and Sample Notes - Panel dimensions and descriptive statistics (selected reported entries preserved): - Panel: Number of panels = 78; Number of periods = 36. - Remittances (overall): Min 3.18, Max 2.94, Observations -0.16, 9.99; N = 2886. - Foreign Aid (overall): Min 11.80, Max 3.22, Observations -13.69, 16.09; N = 2866. - Bank Flows (overall): Min 1.12, Max 9.57, Observations -16.45, 17.07; N = 2886. - Equity Flows (overall): Min 1.32, Max 4.93, Observations -16.23, 17.42; N = 2886. - Geological Disasters (overall): Min 0.30, Max 0.96, Observations 0.00, 14; N = 2808. - Climatic Disasters (overall): Min 1.06, Max 2.10, Observations 0.00, 23; N = 2808. - Human Disasters (overall): Min 0.26, Max 0.71, Observations 0.00, 8; N = 2808. - Low Income Country (LIC) sample and income classification: list of countries and classifications provided in source (e.g., Burundi BDI LIC; Benin BEN LIC; Bangladesh BGD LIC; Ethiopia ETH LIC; Mali MLI LIC; etc.). Note: income group classification based on latest World Bank definition included in source. *IMF Working Paper WP/10/166 — _wp10166 (sections and tables as provided in the source content).* ### Section 1 ### How Do International Financial Flows to Developing Countries Respond to Natural Disasters? ### Abstract and Key Conclusions - Uses multivariate dynamic panel analysis (panel vector autoregression) to examine the response of international financial flows to natural disasters. - Sample: a large sample of developing countries covering the period from 1970 to 2005. - Main empirical findings: - Remittance inflows increase significantly in response to shocks to both climatic and geological disasters. - Responses of aid flows to natural disaster shocks in general tend not to be statistically significant. - International assistance to low income countries increases following geological disaster shocks. - Other private capital flows (bank lending and equity) typically do not attenuate the effects of disasters and in some specifications even amplify the negative economic effects of these events. - Policy implication: - Countries should take their vulnerability to natural disasters into account when considering the costs and benefits of the liberalization of private capital flows. ### Motivation and Objectives (Introduction) - Climate change likely leads to increased variability and more frequent and intense extreme weather events; developing countries are more vulnerable (Hamilton and Fay, 2009). - Prior macroeconomic evidence: - Noy (2009) finds that a one standard deviation increase in the direct damages attributed to a natural disaster could reduce output growth in a developing country by about 9 percent. - Objectives of this paper: - Assess systematically the different responses of development aid, migrant remittances, equity flows and bank lending flows to natural disaster shocks. - Examine whether capital flows exacerbate or attenuate the economic effects of natural disasters. - Inform policy on appropriate responses to expected movements in capital flows following natural disasters and on the role of foreign aid versus private capital. ### Literature Context and Stylized Facts (Links between International Financial Flows and Natural Disasters) - Different types of capital flows exhibit different time-series properties; FDI often considered more resilient than portfolio and bank flows (Kose and others, 2006), though dissenting evidence exists (Claessens, Dooley and Warner, 1995). - Overseas development assistance (ODA) is often volatile and procyclical with adverse macroeconomic implications for aid-dependent countries (Bulir and Hamann, 2008), though results can depend on dataset and aid composition (Hudson and Mosley, 2008). - Remittances often increase or remain stable after large shocks such as natural disasters, macroeconomic crises, and armed conflicts (World Bank, 2005; Yang and Choi, 2007; Yang, 2008). - Contrasting empirical results: - Neagu and Schiff (2009): ODA tends to be more stable than remittances for 73 percent of countries in their sample (1980-2007), but remittances are more stable than FDI. - Raddatz (2007, 2009): aid flows have not significantly attenuated output consequences of natural disasters in his sample. - Yang (2008): hurricane exposure leads to large increases in remittances; foreign aid and remittances seem to increase, while private flows may experience capital flight. - Mohapatra, Joseph and Ratha (2009): micro and macro evidence that remittances increase in aftermath of disasters in countries with a large number of migrants abroad. ### Theoretical Links and Transmission Channels - Neo-classical view: - Destruction of capital stock could raise marginal product of capital and attract capital flows back to steady state. - Countervailing effects: - Disasters that destroy complementary inputs (infrastructure, human capital, public goods) may lower returns to capital, causing capital inflows to be muted or triggering capital outflows. - Disasters can reduce total factor productivity (Loayza and others, 2009), lowering average product of capital and growth prospects, negatively affecting private capital flows (equity, FDI). - Large disasters can induce political instability or weaken rule of law, prompting capital flight. - Market frictions: - If developing countries are capital constrained and international capital markets suffer market failures, reconstruction need may not translate into increased inflows. - Governments and domestic financial intermediaries may borrow abroad (concessional aid, market-priced bank loans, bond issuance) to smooth consumption and finance reconstruction. ### Empirical Approach and Identification (Framework for Analyzing the Response) - Focus on shocks categorized as geological, climatic and human disasters (following Raddatz, 2007). - Main endogenous variables: migrant remittances, foreign aid, bank loans, equity flows; additional controls include interest rate differentials and real exchange rate movements. - Model specification (at country i, time t): - General form: A0 z_it = sum_{j=1}^q A_j z_{i,t-j} + theta_i + epsilon_it (equation (1) in the source) - z_it comprises endogenous variables y_it and weakly exogenous variables x_it. - Identification assumptions: - Natural disaster shocks are assumed to be weakly exogenous and do not respond contemporaneously to shocks in other variables. - Structural identification requires assumptions about the contemporaneous coefficient matrix A0. - Practical estimation choices: - To avoid overparametrization given limited data availability, the number of endogenous variables included in each VAR model was limited and models for each type of capital flow were estimated separately. - Advantages: - Panel VAR methodology allows study of dynamic responses via impulse response functions and forecast error variance decompositions. - Natural disasters can be treated as weakly exogenous, reducing endogeneity concerns. *IMF Working Paper WP/10/166* ### Section 2 ### _wp10166 - Section 2 ### Methodology and Estimation Framework - System: Panel VAR system with endogenous variables CF (specific capital flow: remittances, aid, equity flows or bank loans), idiff (real interest rate differential), ∆RER (change in the real effective exchange rate), Ydiff (output differential) and Nat (natural disaster variable). - Estimation rationale: - Standard panel techniques are feasible, but because some explanatory variables may be endogenous, correlated with fixed effects, and subject to measurement error, the Generalized Method of Moments (GMM) estimator is preferred. - Lagged values of the variables are used as instruments in the GMM specification (models in Section V). - Caveats: GMM is less efficient than some alternatives and may be biased if instruments are weak (reference to Baltagi, 2005). - Dynamic analysis tools: - Impulse response functions are used to simulate dynamic effects of specific natural disaster shocks on capital flows. - Forecast error variance decompositions assess the contribution of different shocks to variability of capital flows at specific horizons. ### Data Coverage and Construction - Sample period: 1970-2005. - Capital flows data sources: - Overseas development assistance, bank and equity flows from the World Bank’s Global Development Finance database. - Remittance inflows to developing countries from the World Bank’s Development Prospects Group. - Natural disaster data source: OFDA/CRED International Emergency Disasters Database (EM-DAT). - Variable construction notes: - Capital flows are expressed in real dollar terms (deflated by the US CPI index). - The log of the real interest rate differential between domestic rates and international rates (e.g., three-month U.S. treasury bill rate) is used to capture the “investment” motive; real ex-post interest rates = nominal interest rate − actual inflation rate observed in the year. - ∆RER = first difference of the log of the real trade-weighted effective exchange rate. - Ydiff = log of income differentials between the developing country and the United States (real GDP per capita PPP adjusted from Penn World Tables). - Natural disaster classification (following Raddatz, 2007): - Climatic events: floods, droughts, extreme temperatures, hurricanes. - Geological events: earthquakes, landslides, volcano eruptions, tidal waves. - Human disasters: famines and epidemics. - EM-DAT inclusion criteria for a disaster event: at least one of (a) 10 or more people killed; (b) 100 or more people affected; (c) a state of emergency declared; (d) a call for international assistance. ### Stationarity and Panel Unit Root Tests (Table 1) - Panel dimensions: Number of panels = 78; Number of periods = 36. Time trends and panel means included. Lag-length chosen by AIC. - Levin-Lin-Chu unit-root test (Adjusted t*): - Remittances: -13.31, p-value 0.00 - ODA: -10.08, p-value 0.00 - Bank: -18.28, p-value 0.00 - Equity: -3.35, p-value 0.00 - Income Differential: -5.17, p-value 0.00 - Δ Real Exchange Rate: -22.01, p-value 0.00 - Interest Differential: -13.23, p-value 0.00 - Breitung unit-root test (lambda): - Remittances: -2.93, p-value 0.00 - ODA: -4.55, p-value 0.00 - Bank: -18.60, p-value 0.00 - Equity: -6.93, p-value 0.00 - Income Differential: 2.66, p-value 0.99 - Δ Real Exchange Rate: -14.62, p-value 0.00 - Interest Differential: -8.41, p-value 0.00 - Hypotheses: Ho: Panels contain unit roots; Ha: Panels are stationary. Common AR parameter. - Im-Pesaran-Shin unit-root test (Z-t-bar): - Remittances: n.a. - ODA: n.a. - Bank: n.a. - Equity: n.a. - Income Differential: -5.27, p-value 0.00 - Δ Real Exchange Rate: -18.79, p-value 0.00 - Interest Differential: -14.93, p-value 0.00 - Hypotheses: Ho: All panels contain unit roots; Ha: Some panels are stationary. Panel specific AR parameter. - Interpretation: - Tests strongly suggest remittance flows, net development aid, net bank loans, net equity flows, the interest rate differential and the first difference of the real effective exchange rate do not follow unit root processes. - Income differential test results are ambiguous: LLC and IPS indicate stationarity, Breitung suggests unit roots. - Overall conclusion: non-stationarity is not a major concern; estimation of Panel VAR models in levels is appropriate. ### Preliminary Fixed-Effects Regression Specification - Dynamic fixed-effects panel model estimated with standard fixed-effect estimator: - Equation form: CF_{it} = Σ lags of CF + Σ lags of Nat + Σ lags of other determinants + country fixed effects + error (see equation (2) in source). - Econometric cautions: - Fixed-effect estimator is biased when lagged endogenous variables are included, but bias decreases with number of time observations. - Alternative estimators (GMM, IV) have drawbacks: less efficiency and potential bias with weak instruments. - Endogeneity concerns exist for some macro variables on the RHS (interest rate and output differentials). Natural disaster variables are assumed weakly exogenous except possibly human disasters (famines, epidemics). - Robust standard errors clustered by countries are reported for all models. ### Main Empirical Findings — Remittances (Table 2 results summarized) - Interpretation: Coefficients for natural disasters are semi-elasticities; the impact elasticity = semi-elasticity × level of disaster variable (formalized in equation (3)). - Climatic and geological disasters: - Remittance inflows tend to increase following climatic and geological disasters. - Quantitative examples: - Remittances typically increase by 0.1 percent in the same time period following an increase of one standard deviation in the incidence of climatic disasters (evaluated at the mean value of the climatic disaster variable across the sample). - For a country at the 95th percentile of climatic disaster incidence, the contemporaneous elasticity to a standard deviation increase ≈ 0.2 percent. - For geological disasters, the elasticity is smaller: for a one standard deviation increase evaluated at within-and-between mean, a country in the 95th percentile of geological disaster incidence would experience an elasticity ≈ 0.1 percent. - Additional lags of disaster variables are typically not statistically significant at conventional levels. - When a deterministic time trend is included, remittances respond only to climatic disasters in a statistically significant way. - When both the deterministic trend and the first difference of the real effective exchange rate are included, disaster variables are no longer statistically significant; note this specification dramatically reduces sample size because RER covers shorter time spans. - Human disasters (famines and epidemics): - No evidence that remittances respond to human disasters in any attempted specifications. - Possible explanation: famines and epidemics may be related to political instability and poor governance, which may discourage additional remittances. ### Main Empirical Findings — Official Development Assistance (Table 3 results summarized) - General result: In most specifications, development aid does not respond to disasters in a statistically significant way. - Where significant: - Aid responds more slowly relative to other financial flows. - Aid increases by around 0.3 percent two years after a one standard deviation increase in geological disasters for a country in the 95th percentile of geological disaster incidence. - Robustness: - Results for aid are generally robust to the inclusion of a deterministic trend and the real effective exchange rate. *Source: _wp10166 - Section 2 (PDF chapter/section), data and tables as provided in the source content.* ### Section 3 ### _wp10166 - Section 3 ### Remittances: response to disasters - Remittances are responsive to natural disasters (both geological and climatic), but not to human disasters. - The elasticities calculated from estimated semi-elasticities indicate that the economic importance of these responses is typically small. - Fixed-effects panel regressions reported in Table 2 (dependent variable: log of remittance inflows) show: - Climatic Disasters coefficients: positive and often statistically significant (examples in Table 2: 0.038***, 0.023***, 0.024**, 0.026**, 0.024**, 0.009 with clustered robust standard errors shown in brackets). - Income Differential coefficients: positive and frequently significant (examples: 0.361**, 0.364*, 0.388**, 0.299**, 0.316**, 0.330**, 0.394**, 0.418**, 0.433***, etc., with corresponding standard errors). - Remittances (t-1) strong persistence: coefficients around 0.817*** to 0.895*** in various specifications (examples: 0.867***, 0.870***, 0.874***, 0.887***, 0.893***, 0.895***; robust standard errors reported). - Sample/estimation notes from Table 2: - Observations: 2450 (in many specifications), with alternative smaller samples reported (1772, 1696, 2371). - R-squared reported around 0.797–0.835 across specifications. - Number of countries: 78 (noted in table). ### Net international development aid (ODA) - In general, aid flows do not present statistically significant responses to natural or human disasters, except for geological disasters where international aid flows seem to respond with substantial delay. - Fixed-effects panel regressions for net foreign aid flows (dependent variable: log of net international development assistance inflows) in Table 3 show: - Climatic Disasters coefficients: generally negative but not statistically significant in listed specifications (examples: -0.054, -0.039, -0.074, -0.058, -0.022, -0.030 with standard errors in brackets). - ODA (t-1) shows strong persistence: 0.458***, 0.457***, 0.457*** and other significant lagged coefficients across specifications. - Some coefficients for Income Differential and Real Interest Rate Differential vary in sign and significance across lags and specifications; geological-disaster related coefficients at lag sometimes show significance (see Table 3 entries for Geological Disasters(t-2): 0.111**, 0.111**, 0.135**, 0.146**). - Sample/estimation notes from Table 3: - Observations typically 2449 with alternative sample sizes (2370, 1772, 1696). - R-squared values around 0.131–0.232 depending on specification. - Number of countries: mostly 78, with some specifications at 76. ### Bank lending flows: negative and short-lived responses - Bank lending flows respond negatively to climatic disasters and to human disasters, with the response tending to occur relatively rapidly and contemporaneously. - Magnitudes reported in main text: - Net bank lending flows typically decrease by around 0.8 percent within the same time period (one year) following a one standard deviation increase in climatic disasters. - For human disasters, bank flows decrease by about 0.1 percent following an increase of one unit in the incidence of disasters. - Table 4 fixed-effects regressions (dependent variable: bank lending flows) key points: - Climatic Disasters coefficients negative and often statistically significant (examples: -0.355**, -0.256*, -0.320**, -0.267*, with standard errors in brackets). - Human Disasters contemporaneous coefficients negative and often significant (examples: -0.638**, -0.629**, -0.568*, -0.549*, -0.437*, -0.480). - Lag structure shows negative effects primarily contemporaneously; additional lags generally do not produce persistent statistically significant negative effects. - Specific disaster types: Landslides show high negative semi-elasticity (examples in Table 4: Landslides -1.388***, -1.046**). - Large Human Disasters coefficient reported: -1.563* (with standard error reported). - Robustness: - Negative effects hold when additional time lags are included, but negative effects are only statistically significant contemporaneously, suggesting non-persistence. - Negative effects remain when the effective real exchange rate is included, but results are not robust when both a deterministic trend and the real exchange rate are included. - Fixed-effects results for bank flow models are noted as not very robust and model fit is relatively poorer. ### Net equity flows: positive lagged responses for some disaster types - Net equity flows increase following climatic and geological disasters, but typically with a one-year lag; responses to human disasters appear with longer lags (two-year lag) and significance depends on lag specification. - Magnitudes and examples provided in main text: - Net equity flows increase by around 0.5 percent one year after a one standard deviation increase in the incidence of climatic disasters. - For a country in the 95th percentile in terms of incidence of climatic disasters, the increase is 1.3 percent one year after a one standard deviation increase. - Net equity flows increase by 0.1 percent one year after a one standard deviation increase in geological disasters. - Equity flows increase by about 0.1 percent two years after a one standard deviation increase in human disaster events. - Robustness: - Results for geological disasters and human disasters are robust to inclusion of the real effective exchange rate. - When both a time trend and the real exchange rate are included, only the response of equity flows to geological disasters remains statistically significant. - Interpretation caveat: many developing countries in the sample do not have access to international equity markets, limiting generalizability. - Conclusion drawn: equity flows behave differently from other private flows and might mitigate macroeconomic consequences of certain disaster events, but robustness varies by disaster type and specification. ### Overall conclusions and methodological notes - Summary of principal findings: - Remittances: responsive to geological and climatic disasters (elasticities small); not responsive to human disasters. - Aid flows: generally no statistically significant responses to disasters, except delayed responses to geological disasters. - Bank lending flows: respond negatively to climatic and human disasters, typically contemporaneously and not persistently; specific events (e.g., landslides, large human disasters) show stronger negative effects. - Equity flows: increase after geological and climatic disasters with a one-year lag and after human disasters with a two-year lag; responses to geological disasters are more robust. - Model and goodness-of-fit observations: - Bank and equity flow models present relatively poorer measures of goodness of fit, suggesting omitted-variable concerns are more prominent for these models. - Dependent variables: logs of respective inflows (remittances, net international development assistance, bank lending flows, equity flows); macroeconomic variables expressed in logs. - Real exchange rate variable used is the first difference of the log of the real exchange rate index. - Standard errors reported are robust and clustered by country; significance notation: *** p<0.01, ** p<0.05, * p<0.1. - Deterministic trends and inclusion of real exchange rate influence robustness of some estimated effects. *Source: _wp10166 - Section 3 (tables and text as provided).* ### Section 4 ### Section 4 ### Fixed-effects panel regressions (equity and bank flows) — key empirical patterns - Dependent variables in panels: log of net bank lending inflows and log of net equity inflows. Robust standard errors clustered by countries were used. All macroeconomic variables expressed in logs. The real exchange rate is the first difference of the log of the real exchange rate index. - Equity flows: lagged equity flows are highly persistent with coefficients such as 0.320***, 0.331***, 0.291***, 0.279***, 0.296***, 0.323*** across specifications (standard errors reported, e.g., [0.042], [0.045], [0.051]). - Bank flows: fixed-effect results (table excerpts) show modest R-squared values in many specifications (examples: 0.177, 0.174, 0.175; alternative specifications 0.113, 0.117, 0.107). - Several disaster-type variables (climatic, geological, human) and their lags were included; some coefficients are statistically significant at conventional levels (e.g., Climatic Disasters(t-1) 0.239** with [0.102]; Geological Disasters(t-1) 0.596** with [0.253]; Earthquake 0.885*** with [0.172]; Earthquake (t-1) 0.589** with [0.237]). - Real exchange rate contemporaneous changes show large and significant coefficients in some specifications (examples: 1.974*** [0.700], 2.021*** [0.712], 2.228*** [0.807]); real exchange rate lags often insignificant. - Deterministic trend is positive and significant in many regressions (examples: 0.062*** [0.020], 0.076*** [0.020], 0.061*** [0.019]). ### PVAR analysis — remittance inflows - Methodology notes: - Systems estimated using GMM-panel VAR (PVAR) with one and two lags depending on sample and variable; forward mean differencing (Helmert transformation) applied to address correlation between country fixed effects and lagged dependent variables. - 10 countries excluded from VAR estimation due to missing years: Benin, Buthan, Guinea, Guinea-Bissau, Equatorial Guinea, Lebanon, Liberia, Nicaragua, Paraguay and Tunisia. - Impulse response findings (remittances): - Climatic disaster shocks: remittances increase on impact; effects persist for about one year and become statistically insignificant thereafter for the full sample. - Forecast error variance decomposition (FEVD): climatic disaster shocks responsible for about 16 percent of the variance of remittance inflows at the 10 year horizon (full sample). - Geological disaster shocks: remittances present a positive, statistically significant response on impact that persists after a year and remains marginally significant even two years after the shock (full sample). - FEVD: geological disasters account for 34 percent of the variance in remittances for this model (full sample). - Large-disaster subsample (both climatic and geological): similar conclusions reported (impulse responses not shown in source to save space). - Low-income countries (LICs) subset: - Climatic shocks: remittance increase on impact and response in LICs is more persistent than full sample, remaining statistically significant even after two periods. - FEVD for LICs: climatic shocks seem to account for over 50 percent of the FEVD for remittances. - Geological shocks in LICs: contrary to full sample, remittance inflows present a statistically significant increase only on impact. - Overall: PVAR results imply the economic significance of remittance responses to natural disasters is higher than suggested by fixed-effects regressions. ### PVAR analysis — international aid (ODA) flows - Full sample: - No statistically significant response of international aid to climatic shocks. - Response of aid to geological shocks is negative but barely statistically significant; error bands are wide. - Large-disaster models: response of foreign aid to geological shocks not statistically significant; response to large climatic shocks is significant on impact, but FEVD shows climatic shocks account for merely 1 percent of variance in aid flows. - Low-income countries (LICs): - Models suggest aid increases on impact following a geological disaster shock. - Response of aid to climatic shocks in LICs is not statistically significant. - FEVD: geological events account for 11 percent of the variance of aid flows to low income countries. - Interpretation: the statistically significant response of aid to geological shocks in LICs indicates foreign aid can attenuate negative impacts of geological disasters in the poorest group, while aid’s role following climatic shocks is limited in the full, heterogeneous sample. ### PVAR analysis — net bank lending flows - Sample adjustments: models exclude Brazil, India and China because responses of bank flows in those large emerging markets are atypical and exclusion improved model performance and narrowed error bands. - Impulse response findings (bank flows): - Bank flows present a negative and significant response to climatic and geological disasters on impact; this negative response generally becomes statistically indistinguishable from zero in subsequent periods. - FEVD (Table 7 summary in text): geological shocks account for 23 percent of the variation in bank flows at the 10 period forecast horizon; climatic shocks account for only 2 percent of the variation. - Large-disaster models: when only large climatic disasters are considered, the response of bank flows is not statistically significant. For large geological disasters, the response is significant and positive on impact, but FEVD shows large geological disasters account for less than one percent of the variance in bank lending flows. - LICs subset: - Bank flow responses follow a similar pattern to the full sample, with an exception: the response to geological shocks shows a marginally significant increase after one year in LICs. ### Synthesis — relative roles of flows in disaster episodes - Remittances: - Increase on impact to both climatic and geological shocks; more persistent and economically important for remittances than preliminary fixed-effects estimates suggested. - FEVD indicates substantial shares of remittance variance attributable to disaster shocks (full sample: climatic 16 percent, geological 34 percent; LICs: climatic >50 percent). - International aid: - Limited response to climatic shocks in the diverse full sample; more targeted and meaningful response to geological shocks in LICs (geological events explain 11 percent of ODA variance in LICs). - Bank lending: - Immediate negative impact on bank flows from both climatic and geological shocks in the short run; geological shocks have greater importance for variation in bank lending over a 10-period horizon (23 percent) than climatic shocks (2 percent). - Implication: different external financing sources react differently to disaster types and across country income groups. Remittances are a major immediate and, in some samples, persistent offset following disasters; aid responds more to geological disasters in LICs; private bank flows contract on impact but generally lack persistence. *Source: _wp10166 - Section 4* ### Section 5 ### Section 5 ### Impulse response and short-run dynamics of equity and bank flows - For the full sample of developing countries, net equity flows increase on impact following a climatic disaster shock, but the response becomes statistically insignificant in subsequent periods. - Equity flows respond relatively rapidly to climatic disasters, but the effects on these financial flows tend to be short-lived. - The impulse responses of equity flows to geological disaster shocks are not statistically significant. - When only large natural disasters are included in the models, the response of equity flows is not statistically significant. - For low-income countries (LICs), equity flows do not present statistically significant responses to either climatic or geological shocks. - Bank flows do not tend to attenuate the impact of natural disaster shocks and in most specifications are likely to compound their negative economic impacts; in some specifications net bank lending outflows occur after the onset of disasters, amplifying negative economic effects. ### Forecast Error Variance Decomposition (FEVD) — key numbers (at t=10) - Forecast Error Variance Decomposition for Bank flows (variance of bank flows explained by shock in each variable) - Climatic Events: 0.02 n.a. 0.13 n.a. - Income Differential: 0.02 0.03 0.01 0.01 - Real Exchange Rate: 0.00 0.00 0.00 0.00 - Interest Differential: 0.01 0.01 0.00 0.00 - Bank: 0.95 0.73 0.86 0.94 - Geological Events: n.a. 0.23 n.a. 0.05 - Forecast Error Variance Decomposition for Equity flows (variance of equity flows explained by shock in each variable) - Climatic Events: 0.29 n.a. 0.00 n.a. - Income Differential: 0.03 0.01 0.01 0.01 - Real Exchange Rate: 0.02 0.01 0.00 0.00 - Interest Differential: 0.00 0.00 0.00 0.00 - Equity: 0.66 0.93 0.99 0.99 - Geological Events: n.a. 0.05 n.a. 0.00 - Climatic disasters account for 13 percent of the forecast error variance of bank flows for low-income countries, whereas geological disasters account for only 5 percent (as noted in the text). - Climatic shocks account for about 29 percent of the forecast error in equity flows to the developing countries considered in the sample (full sample). - Natural disasters can account for up to 53 percent of the forecast error variance of remittance inflows to low-income countries (from conclusions). ### Main conclusions on different flow types - Remittances: - Increase significantly in response to shocks to both climatic and geological disasters. - Natural disasters can account for up to 53 percent of the forecast error variance of remittance inflows to low-income countries. - Remittances exhibit a compensatory nature for disaster-affected countries. - Foreign aid: - International aid to low-income countries increases following geological disaster shocks. - In more general circumstances, aid flows’ responses to natural disaster shocks are not statistically significant. - International aid typically plays a limited role in attenuating economic consequences of disasters, but its role is more significant in poorer countries. - Bank lending: - Bank lending flows generally do not attenuate the effects of disasters and can amplify negative economic effects via net bank lending outflows after disasters. - Equity flows: - Not an important source of finance for disaster recovery in low-income countries due to shallow equity markets. - Respond positively to climatic disasters for the larger sample of developing countries (excluding Brazil, China, India), but not to geological disaster shocks; the positive effect is short-lived. ### Methodological caveats and avenues for future research - Scope to refine natural disaster measures to introduce additional variation in the magnitude of economic impact. - Further differentiation between types of financial flows is desirable (e.g., budget support versus project-related international assistance). - The GMM estimator used for the PVAR models assumes all countries in the sample follow the same dynamics; coefficient estimates will be biased if this is not the case. - Subsequent work could test results with alternative estimators not subject to the same problem (such as mean-group estimators). ### Policy implications and recommendations - Foster remittance flows for countries vulnerable to natural disasters, for example by pursuing policies to reduce transaction costs associated with remittances. - Policymakers in disaster-stricken countries typically should not rely on substantial increases in foreign assistance to finance reconstruction or consumption smoothing needs, except in specific circumstances. - The international community should continue efforts to strengthen capacity to scale-up foreign aid following natural disaster events, given that historically recorded financial assistance provided only a mitigated response, particularly for climatic disasters. - Consider the absorptive capacity of recipient countries when scaling up aid; donor concerns about absorptive capacity may partly explain limited aid responses. - In capital/financial account management, countries should carefully consider their vulnerability to natural disasters and the impact of disasters on capital flows when weighing benefits and costs of capital account liberalization. - Benefits of liberalization in facilitating financing for disaster recovery and consumption smoothing typically do not appear to have materialized for many developing countries. - There is evidence of considerable risks of private capital outflows following natural disaster events. *Source: _wp10166 - Section 5* ### Section 6 ### Section 6 ### Geological Disasters - overall: Min 0.30, Max 0.96, Observations 0.00, 14; N = 2808 - overall: Min 0.10, Max 0.39, Observations 0.00, 4.00; N = 1116 - between: Min 0.66, Max 0.00, Observations 3.69; n = 78 - between: Min 0.22, Max 0.00, Observations 0.86; n = 31 - within: Min 0.70, Max -3.39, Observations 10.61; T = 36 - within: Min 0.33, Max -0.76, Observations 3.30; T = 36 ### Climatic Disasters - overall: Min 1.06, Max 2.10, Observations 0.00, 23; N = 2808 - overall: Min 0.76, Max 1.36, Observations 0.00, 11.00; N = 1116 - between: Min 1.63, Max 0.00, Observations 8.89; n = 78 - between: Min 0.96, Max 0.08, Observations 5.44; n = 31 - within: Min 1.34, Max -7.83, Observations 15.17; T = 36 - within: Min 0.97, Max -4.68, Observations 6.32; T = 36 ### Human Disasters - overall: Min 0.26, Max 0.71, Observations 0.00, 8; N = 2808 - overall: Min 0.42, Max 0.89, Observations 0.00, 8.00; N = 1116 - between: Min 0.30, Max 0.00, Observations 1.64; n = 78 - between: Min 0.26, Max 0.11, Observations 1.25; n = 31 - within: Min 0.65, Max -1.38, Observations 7.01; T = 36 - within: Min 0.85, Max -0.83, Observations 7.17; T = 36 ### Income Differential - overall: Min 2.28, Max 0.88, Observations -0.84, 4.48; N = 2886 - overall: Min 1.51, Max 0.55, Observations -0.84, 2.93; N = 1147 - between: Min 0.82, Max 0.81, Observations 4.21; n = 78 - between: Min 0.41, Max 0.81, Observations 2.35; n = 31 - within: Min 0.31, Max 0.42, Observations 4.29; T = 37 - within: Min 0.38, Max -0.34, Observations 2.95; T = 37 ### Real Exchange Rate - overall: Min -0.02, Max 0.31, Observations -11.66, 2.20; N = 1942 - overall: Min -0.03, Max 0.18, Observations -1.88, 1.00; N = 762 - between: Min 0.03, Max -0.24, Observations 0.04; n = 76 - between: Min 0.02, Max -0.10, Observations 0.02; n = 30 - within: Min 0.31, Max -11.45, Observations 2.41; T- bar = 25.5 - within: Min 0.18, Max -1.83, Observations 1.04; T-bar = 25.4 ### Interest Differential - overall: Min -0.03, Max 0.24, Observations -4.58, 1.67; N = 2604 - overall: Min -0.05, Max 0.26, Observations -4.58, 0.52; N = 1054 - between: Min 0.12, Max -0.78, Observations 0.18; n = 78 - between: Min 0.15, Max -0.78, Observations 0.08; n = 31 - within: Min 0.22, Max -3.94, Observations 1.63; T- bar = 33.4 - within: Min 0.22, Max -3.86, Observations 0.78; T-bar = 34 ### Remittances - overall: Min 3.18, Max 2.94, Observations -0.16, 9.99; N = 2886 - overall: Min 2.17, Max 2.41, Observations -0.16, 8.53; N = 1147 - between: Min 2.23, Max 0.00, Observations 8.38; n = 78 - between: Min 1.90, Max 0.00, Observations 6.58; n = 31 - within: Min 1.93, Max -3.94, Observations 8.46; T = 37 - within: Min 1.52, Max -4.42, Observations 7.45; T = 37 ### Foreign Aid - overall: Min 11.80, Max 3.22, Observations -13.69, 16.09; N = 2866 - overall: Min 12.63, Max 1.06, Observations 3.66, 16.09; N = 1144 - between: Min 2.24, Max 0.00, Observations 14.69; n = 78 - between: Min 0.80, Max 10.99, Observations 14.27; n = 31 - within: Min 2.33, Max -13.01, Observations 20.17; T- bar = 36.7 - within: Min 0.72, Max 5.16, Observations 16.42; T-bar = 36.9 ### Bank Flows - overall: Min 1.12, Max 9.57, Observations -16.45, 17.07; N = 2886 - overall: Min -0.40, Max 7.63, Observations - 14.43, 15.26; N = 1147 - between: Min 3.26, Max -5.26, Observations 10.41; n = 78 - between: Min 1.87, Max -5.26, Observations 3.62; n = 31 - within: Min 9.00, Max -25.07, Observations 21.64; T = 37 - within: Min 7.41, Max - 16.58, Observations 20.12; T = 37 ### Equity Flows - overall: Min 1.32, Max 4.93, Observations -16.23, 17.42; N = 2886 - overall: Min 1.03, Max 3.38, Observations - 11.76, 14.19; N = 1147 - between: Min 1.92, Max -1.48, Observations 9.34; n = 78 - between: Min 1.48, Max 0.00, Observations 5.62; n = 31 - within: Min 4.54, Max -21.04, Observations 15.97; T = 37 - within: Min 3.05, Max - 15.30, Observations 11.92; T = 37 ### Low Income Country Sample — List of Countries Included in the Regressions and their Income Classification - Note: LIC=Low-income country; LMC= Lower Middle-income country; UMC= Upper Middle Income country, HIC High Income Country. Income group classification based on latest World Bank definition. - Argentina ARG UMC - Burundi BDI LIC - Benin BEN LIC - Burkina Faso BFA LIC - Bangladesh BGD LIC - Bolivia BOL LMC - Brazil BRA UMC - Barbados BRB HIC - Bhutan BTN LMC - Botswana BWA UMC - Chile CHL UMC - China CHN LMC - Cote d'Ivoire CIV LIC - Cameroon CMR LMC - Congo, Re p. COG LMC - Colombia COL LMC - Cape Verde CPV LMC - Costa Rica CRI UMC - Algeria DZA LMC - Ecuador ECU LMC - Egypt, Arab Rep. EGY LMC - Ethiopia ETH LIC - Fiji FJI UMC - Gabon GAB UMC - Ghana GHA LIC - Guinea GIN LIC - Gambia, The GMB LIC - Guinea-Bissau GNB LIC - Equatorial Guinea GNQ HIC - Guatemala GTM LMC - Guyana GUY LMC - Honduras HND LMC - Indonesia IDN LMC - India IND LMC - Jamaica JAM UMC - Jordan JOR LMC - Kenya KEN LIC - Lebanon LBN UMC - Lesotho LSO LMC - Morocco MAR LMC - Madagascar MDG LIC - Mali MLI LIC - Malawi MWI LIC - Malaysia MYS UMC - Mexico MEX UMC - Niger NER LIC - Nigeria NGA LIC - Nicaragua NIC LMC - Nepal NPL LIC - Pakistan PAK LIC - Peru PER LMC - Philippines PHL LMC - Papua New Guinea PNG LIC - Paraguay PRY LMC - Rwanda RWA LIC - Senegal SEN LIC - Sierra Leone SLE LIC - Swaziland SWZ LMC - Seychelles SYC UMC - Syrian Arab Republic SYR LMC - South Africa ZAF UMC - Tanzania TZA LIC - Thailand THA LMC - Trinidad and Tobago TTO HIC - Tunisia TUN LMC - Turkey TUR UMC - Uganda UGA LIC - Uruguay URY UMC - Venezuela, RB VEN UMC - Zambia ZMB LIC - Zimbabwe ZWE LIC - Chad TCD LIC - Togo TGO LIC - Liberia LBR LIC - Congo, Dem. Rep. ZAR LIC - Somalia and other country entries appear in the list as provided above. *Source: _wp10166 - Section 6* --- _Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10166.pdf_