## _wp0585

## Source details

**Canonical URL:** [_wp0585](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0585.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0585.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0585.pdf.json)

---

### Introduction and research question
- Research question: Does the choice of exchange rate regime affect an economy’s adjustment to real shocks (natural shocks such as wind storms—hurricanes, tornados, typhoons—and earthquakes)?
- Theoretical background:
  - Nominal exchange rate movements can restore equilibrium faster when prices are rigid (Friedman, 1951; Mundell, 1961; Poole, 1970).
  - Fixed rate regimes can reduce exchange rate variability and lower transaction costs, potentially stimulating trade, investment and growth (Frankel and Rose, 2002).
  - Credibility and central bank capacity to pursue countercyclical monetary policy may be limited in many developing countries (Kaminsky et al., 2004; Schmukler and Serven, 2000).
- Identification advantage: Natural shocks (windstorms and earthquakes) are mostly low probability random events whose incidence is unrelated to policy choices.

### Data, measurement, and empirical framework
- Data and sample:
  - Data on windstorms and earthquakes from the Center for the Research on the Epidemiology of Disasters (CRED).
  - CRED records a disaster when any of: "10 or more reported killed; 100 people reported affected; a call for international assistance; a declaration of a state of emergency."
  - Sample: 120 countries over 1961-2000.
  - In the most inclusive specification: "240 reported earthquakes and 423 reported windstorms".
- Key variables:
  - Investment: "The Ratio of Gross Capital Formation, net of Inventories, to GDP." Source: World Bank (2003).
  - FDI: "The Ratio of Foreign Direct Investment to GDP." Source: World Bank (2003).
  - GDP Growth: "Growth in Real GDP." Source: World Bank (2003).
  - Exchange rate regimes: Reinhart and Rogoff (2004) 6-way annual classification used to construct binary fixed/flexible regime variable; alternate results reported using six-category classification and Levy-Yeyati and Sturzenegger (2004).
  - Med: indicator = 1 if a country maintained a fixed exchange rate system for at least 64 percent of years from 1960-2000.
- Empirical specification:
  - Outcome y_it regressed on shock indicator S_it, regime indicator R_it, interactions (S*R) and lags up to two years, controls X_it, country fixed effects ε_i, year effects ν_t, residual u_it.
  - Parameters γ_j on interaction terms test whether shock impacts depend on prevailing exchange rate regime.
  - Residuals allowed to be correlated across years for the same country; Huber-White robust standard errors clustered at country level.

### Main empirical findings — Investment (short run)
- Developing countries (DC):
  - Windstorms generate a cumulative increase in investment concentrated in DCs with fixed exchange rate regimes.
  - Pooled three-year horizon: a 3.21 percentage point of GDP increase in the investment level (p-value=0.06).
  - Baseline DC results (Table 3a, column (4)):
    - S_it * R_it: 0.0187** (0.009).
    - S_{t+1} * R_{t+1}: 0.013** (0.006).
    - In the year of the shock: mean investment increases by about 1.87 percentage points of GDP (p-value=0.05) in fixed rate DCs.
    - One year after the shock: investment remains about 1.34 percentage points higher (p-value=0.04) in fixed rate DCs.
    - Two years after the shock: no significant impact.
  - “Constant” regime subsample (regime unchanged over four-year window): S_{t+1} * R_{t+1} = 0.019*** (0.007) in DCs.
  - Descriptive evidence: mean investment level when windstorms occur is "over 2.5 percentage points of GDP (p-value=0.00) higher" in DCs with fixed regimes; no significant response in flexible DCs.
- Advanced countries (AC):
  - No significant investment response regardless of regime.
- Aggregate notes:
  - The cumulative increase in investment over the three years remains around three percentage points of GDP in the basic results.
  - S_it+1*R_it+1 in pooled contexts and subgroup analyses reported with exact coefficients and standard errors as above.

### Main empirical findings — FDI
- Pooled sample (AC + DC, Table 6, Column 2):
  - FDI response positive and significant only for fixed rate regimes in the year of the shock (p-value=0.05) and one year after (p-value=0.06).
- Developing countries:
  - FDI response limited to the year of the shock and significant only in fixed rate regimes.
  - FDI increases by about 0.51 percentage points of GDP (p-value=0.10) in the year of the shock for DCs in fixed regimes.
- Qualitative conclusion: FDI response parallels gross capital formation response.

### Main empirical findings — Growth (short run and dynamics)
- Pooled sample (both AC and DC):
  - Growth about 0.74 (p-value=0.07) percentage points higher in the year of the shock (S_t).
  - Growth about 0.68 (p-value=0.02) percentage points higher in the year immediately afterwards (S_{t+1}).
  - No significant impact beyond one year in the pooled sample.
- Advanced countries:
  - Overall impact small.
  - Selected Table 4a (AC column (3)) coefficients:
    - S_t: -0.175 (0.188).
    - S_{t+1}: 0.726*** (0.254).
  - Evidence that growth rebounds faster in ACs with flexible regimes.
- Developing countries:
  - Cumulative three-year impact economically large and slightly higher in fixed regimes: 1.51 percentage points in fixed regimes versus 1.30 percentage points in flexible regimes.
  - Dynamics differ by regime:
    - Fixed regimes: contemporaneous decline in growth, followed by a large rebound one to two years after the shock. Example DC coefficients (Table 4a, column (4)):
      - S_t: 1.295** (0.602).
      - S_t * R_t: -1.473** (0.611).
      - S_{t+1}: 0.637 (0.425).
      - S_{t+2} * R_{t+2}: 1.687*** (0.536).
    - Flexible regimes: output adjusts faster; nearly all growth increases occur in the year of the shock.

### Mechanisms linking regime choice to adjustment
- Exchange rate uncertainty channel:
  - Import content of investment goods in subset of DCs about "30 percent" (Burstein et al., 2004); exchange rate volatility affects domestic price of imported investment goods.
  - Table 10 (parallel rate volatility):
    - A one standard deviation increase in the parallel rate in the year before the shock is associated with a 2.25 percentage point decline in investment one year after the shock in DCs (text statement).
    - Reported coefficient: S_{t+1} * PSD_{t-1} * 10-04 for DC: -0.273** [0.142].
    - Observations: 645 (AC) and 1327 (DC); R-squared: 0.72 (AC) and 0.59 (DC).
- Export channel and timing:
  - Using exports/GDP reveals export behavior accounts for much of the difference between fixed and flexible regimes.
  - Joint significance: Sit*Rit and Sit*Exports jointly significant at five percent (p-value=0.04) in year of shock — suggests flexible-regime output responses reflect export increases.
  - Empirical magnitudes:
    - Export decline in fixed regimes: 9.45 percentage points of GDP (p-value=0.02) over two years.
    - Real effective exchange rate depreciation in flexible DCs one year after shock: 3.52 percentage point (p-value=0.05).
    - Depreciation in fixed regimes is 44 percent smaller (p-value=0.03).
    - Parallel nominal exchange rate depreciation in flexible regimes: 3.17 percentage points one year after shock.
    - Parallel nominal exchange rate depreciation in fixed regimes: 0.73 percentage point (p-value=0.09).
  - No significant impact on CPI inflation one year after the shock in reported specifications.
- Summary of short-run adjustment paths:
  - Flexible regimes: faster rebound in growth, partly via currency depreciation and higher exports.
  - Fixed regimes: immediate decline in output followed by rebound two years later; adjustment characterized by higher investment in first two years rather than export-led recovery.

### Robustness, regional heterogeneity, and alternative shocks
- Robustness checks:
  - Results robust to allowing shock impact to vary by physical size, island status, per capita income, ratio of aid to GNI, and frequency of storm incidence.
  - “Constant” regime subsample (regime unchanged over four-year window) yields similar results.
- Regional findings (Table 5):
  - Sub-Saharan Africa:
    - One year after windstorm, investment about 5.30 percentage points higher in fixed regimes (Table 5, Columns 2 and 3).
    - No significant contemporaneous growth impact; two years after shock growth lower in flexible regimes (p-value=0.07).
  - Developing Asia:
    - No significant investment response detected.
    - Two years after shock, growth significantly higher in fixed regimes; indication of lower growth in year of shock for fixed regimes (p-value=0.11).
  - Caribbean:
    - Similar responses to overall sample but much larger point estimates; small-sample issues and missing data noted.
- Earthquakes (Table 7):
  - Investment response qualitatively similar to windstorms but smaller in magnitude.
  - DCs with fixed regimes: investment about 1.20 percentage points of GDP higher compared with flexible regimes.
  - DCs with flexible regimes: mean decline in investment about 0.97 percentage points of GDP the year after an earthquake.
  - Growth estimates similar directionally but not statistically significant in DCs; ACs show growth lower in fixed regimes by about 0.88 percentage points (p-value=0.02).
- Different time periods (Table 8):
  - Reporting of windstorms increased after 1980; separate estimations for 1960-80 and 1981-2001 show:
    - Estimates from 1960-80 generally small and often insignificant.
    - Over 1980-2001: cumulative impact in DCs with fixed regimes is 5.42 percentage points of GDP; no significant impact in developed countries or DCs with flexible regimes.

### Long-run effects (1961-2001 averages)
- Long-run analysis using averages and indicator for predominantly fixed regimes (fixed in >64 percent of years):
  - Robust evidence (p-value=0.01) that windstorms positively affect average investment in economies with predominantly fixed regimes.
  - In predominantly fixed regimes, a one standard deviation increase in average number of windstorms associated with a 2 percentage point increase in average investment over the period.
  - Mean frequency of windstorms has no significant impact on investment in economies that were flexible more than 36 percent of years.
  - A one standard deviation increase in average number of windstorms associated with a 0.78 percentage point increase in average growth (p-value=0.06) only in predominantly fixed regimes.
- Threshold and sample notes:
  - Threshold for “mostly fixed” = 64 percent of years (sample median).
  - Columns estimated separately for above/below median show similar patterns.

### Endogeneity, selection concerns, and robustness to regime choice
- Potential selection bias:
  - Severe windstorms could force regime changes (e.g., fixed → flexible), biasing estimates if flexible regimes are correlated with severe storms.
  - Authors restrict to “constant” regimes (columns 5 and 6 of Table 3a) to reduce such bias; results change only slightly.
  - Controlling for frequency of past shocks and physical observables mitigates concerns; lack of data on storm intensity remains a caveat.

### Key statistics and sample sizes (selected, preserved exactly)
- Table 3a NOBs (observations):
  - Pooled Sample: 2843
  - AC (column 3): 748
  - DC (column 4): 2095
  - “Constant” Regimes (AC) (column 5): 712
  - “Constant” Regimes (DC) (column 6): 2013
- Table 3a R-squared examples:
  - Pooled Sample: 0.55
  - AC: 0.67
  - DC: 0.58
- Table 3b NOBs:
  - Typically 2095 observations for most DC specifications; 2020 in columns involving foreign aid.
  - R-squared: 0.61 across reported DC specifications.
- Table 4a NOBs and R-squared:
  - Pooled Sample (column 2): NOBs 1954, R-squared 0.25
  - AC (column 3): NOBs 640, R-squared 0.44
  - DC (column 4): NOBs 1314, R-squared 0.27
  - “Constant” Regimes (AC) (column 5): NOBs 611, R-squared 0.44
  - “Constant” Regimes (DC) (column 6): NOBs 1268, R-squared 0.33

### Policy implications and interpretation
- Flexible exchange rate regimes:
  - Provide faster output adjustment to adverse real shocks via currency depreciation and export responses; beneficial for short-run macroeconomic adjustment speed.
- Fixed exchange rate regimes:
  - Tend to stimulate higher investment responses to shocks—particularly in the first two years after a shock—which can translate into higher average investment and growth over decades for economies that predominantly maintain fixed regimes.
  - May forgo faster short-run adjustment but gain long-run productivity benefits from investment-led recovery.
- Net interpretation:
  - Both views have merit: flexible regimes facilitate faster short-run adjustment via exchange rate movements and the export channel; fixed regimes encourage investment responses that yield long-run growth benefits.
  - Policy trade-off: faster short-run adjustment under flexible regimes versus potential longer-run growth gains under fixed regimes.

*Italic: _wp0585 — robustness, regional and shock-type analyses of exchange rate regime effects on investment, FDI, and growth after windstorms and earthquakes.*

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

### _wp0585 - References..............................................................................................................

### Introduction
- Research question: Does the choice of exchange rate regime affect an economy’s adjustment to real shocks (natural shocks such as wind storms—hurricanes, tornados, typhoons—and earthquakes)?
- Theoretical background:
  - Standard arguments (Friedman, 1951; Mundell, 1961; and Poole, 1970) imply nominal exchange rate movements can restore equilibrium faster when prices are rigid.
  - Fixed rate regimes can reduce exchange rate variability and lower transaction costs, potentially stimulating trade, investment and growth (Frankel and Rose, 2002).
  - Credibility and central bank capacity to pursue countercyclical monetary policy may be limited in many developing countries (Kaminsky et al., 2004; Schmukler and Serven, 2000).
- Identification advantage: Natural shocks (windstorms and earthquakes) are mostly low probability random events whose incidence is unrelated to policy choices, reducing sample selection bias in studying regime effects.

### Main empirical findings
- Investment response (developing countries):
  - Among developing countries with fixed exchange rate regimes, the adjustment process is characterized by an increase in investment of three to five percentage points of GDP over a three year window after the shock.
  - In developing countries with flexible rate arrangements, there is no detectable impact on investment.
- Investment response (advanced economies):
  - In advanced economies, there is no significant investment response regardless of the regime type.
- Growth response (developing countries):
  - The cumulative effect of the shock on growth is about 1.54 percentage points over a three year horizon and is similar across both types of regimes.
  - Timing differs by regime:
    - Flexible regimes: output adjusts much faster; nearly all growth increases occur in the year of the shock.
    - Fixed regimes: contemporaneous decline in growth, followed by a large rebound one to two years after the shock.
- Growth response (advanced economies):
  - Little evidence that the choice of exchange rate regime has an economically meaningful impact on the growth response.

### Interpretation and mechanisms
- Support for both sets of theories:
  - Fixed regimes: exchange rate stability appears to confer important advantages for investment and growth in developing countries by reducing exchange rate uncertainty that can impede investment (noting high import content of investment goods in many developing countries).
  - Flexible regimes: allow depreciation to handle adverse real shocks, temporarily spurring exports and growth and enabling faster adjustment.
- Long-run tradeoff:
  - Over a 30-year period, a higher average incidence of natural shocks is associated with both higher investment and growth only in economies that predominantly maintained fixed regimes.
  - While depreciation can help adjust quickly to adverse shocks, exchange rate variability itself seems to impede the investment response to these shocks.
  - Economies that invest in response to shocks appear to experience faster economic growth from natural shocks in the long run.

### Robustness and identification notes
- Results are robust to various modifications and sensitivity analyses.
- The use of natural shocks helps mitigate concerns about regime choice being endogenous to shock incidence; susceptibility to natural shocks can be controlled using physical observables in estimation.

### Organization of the paper (as presented)
- Section II: data discussion.
- Section III: main results.
- Section IV: sensitivity analyses.

* _wp0585 - References (PDF chapter/section) *

### Section V discussion.

### Section V discussion

### Data
- Source and scope:
  - Data on windstorms and earthquakes are taken from the Center for the Research on the Epidemiology of Disasters (CRED).
  - The CRED database records a disaster when any of the following conditions occur: "10 or more reported killed; 100 people reported affected; a call for international assistance; a declaration of a state of emergency."
  - Sample: 120 countries over the period 1961-2000.
  - In the most inclusive specification there are "240 reported earthquakes and 423 reported windstorms" in the sample.
- Key observations on economic impact and mechanisms:
  - Natural shocks both directly and indirectly affect the capital stock (example: Hurricane Mitch struck Honduras in 1998, causing severe unemployment in urban areas since the disaster affected business capital).
  - Developing countries often rely on imported building materials and equipment to rebuild, with long term consequences (example: Bangladesh after 1987–1988 storms relaxed restrictions on private agricultural investment and imports of equipment; access to new technologies was associated with rapid expansion of dry-season irrigated rice, displacing flood-prone deep water rice and jute, greatly reducing the volatility and level of the price of rice).
  - The choice of exchange rate regime can potentially shape the economic adjustment to real shocks.

### Measurement and variables
- Exchange rate regime measurement:
  - Principal measure: Reinhart and Rogoff (2004) 6-way annual exchange rate classification, used to construct a binary variable denoting “fixed” or “flexible” regimes.
  - Also report results using the six-category classification itself and the Levy-Yeyati and Sturzenengger (2004) method.
- Key economic variables and definitions (selected from Table 2):
  - Investment: "The Ratio of Gross Capital Formation, net of Inventories, to GDP." Source: World Bank (2003).
  - FDI: "The Ratio of Foreign Direct Investment to GDP." Source: World Bank (2003).
  - GDP Growth: "Growth in Real GDP." Source: World Bank (2003).
  - Frequency: "The Percent of Years With At Least One Recorded Windstorm, 1900-2000." Source: CRED (2003).
  - Exchange Rate Dummy: "1 if Exchange Rate Regime Classification (RR) 3≤; 0 otherwise. Note '5' or 'freely falling' are excluded from the sample." Source: Reinhart and Rogoff (2004).
  - Med: "An indicator variable that takes on the value of 1 if a country maintained a fixed exchange rate system—see Exchange Rate Dummy—for at least 64 percent of the years from 1960-2000: the median percent among developing countries in the sample." Author’s Calculations.
  - Control variables include per capita income, trade openness, population density, urbanization, population size, land area, government consumption growth, terms of trade growth, exports, aid, and democracy score.

### Empirical framework
- Theoretical motivation:
  - A simple two-period model where a representative firm chooses investment I in period one before learning investment cost shocks.
  - A multiplicative shock a exponentially affects investment cost: total cost in period one is 2a e I (as described in text); a is normally distributed with mean 0 and variance 2σ.
  - Production uses Cobb-Douglas technology; depreciation parameter δ captures the adverse impact of a shock on the capital stock.
  - Comparative statics:
    - By reducing the capital stock, shocks can increase the optimal level of investment in period one: "(), 0 I∂δσ > ∂δ."
    - But the investment response to an increase in δ negatively depends on the level of uncertainty surrounding the cost of investment: "(), 22 2 ,0. I∂δσ < ∂δ∂σ."
- Empirical specification (equation (1)):
  - Let Sit = indicator that equals one if a shock occurs in country i in year t.
  - Let Rit = indicator that equals one if the exchange rate regime is classified as fixed, zero otherwise.
  - Outcome variable yit (e.g., investment, FDI, real GDP growth) is regressed on interactions of S and R, S and R individually, control vector Xit, country fixed effects εi, year effects vt, and residual uit.
  - The parameters γj on the interaction terms (SR and lags) test whether the impact of a shock on the outcome depends on the prevailing exchange rate regime.
  - Effects of shocks are considered in the current year and up to two years after the shock.
  - Time-varying controls include population density, urbanization, per capita income, and trade openness.
  - Residuals are allowed to be correlated across years for the same country.

### Main empirical results — Investment
- Descriptive evidence (Figure 1 preview):
  - For advanced countries: the mean difference in the investment level (ratio of gross capital formation net of inventories to GDP) between years with and without windstorms is small, not statistically different from zero, and does not depend on exchange rate regime.
  - For developing countries with fixed exchange rate regimes: the mean investment level when windstorms occur is "over 2.5 percentage points of GDP (p-value=0.00) higher" than the mean level in non-windstorm years.
  - For developing countries with flexible rate regimes: there is no significant investment response.
- Regression results (Table 3a, baseline specification controlling for per capita income, trade, population density, urbanization, year and country fixed effects):
  - When advanced and developing countries are combined: windstorms have no significant impact on investment, and the choice of exchange rate regime does not influence adjustment.
  - Disaggregated sample — developing countries only:
    - In the year of the shock (column 4): "the mean level of investment in developing countries operating under fixed rate regimes increases by about 1.87 percentage points of GDP (p-value=0.05) when windstorms occur."
    - One year after the shock: the investment level remains about "1.34 percentage points higher (p-value=0.04) in fixed rate economies."
    - Two years after the shock: "There is no significant impact two years after the shock."
  - The text notes a cumulative impact is calculated, but the reported excerpt ends before the cumulative figure is provided.

### Interpretation and mechanisms emphasized
- Exchange rate uncertainty as a transmission channel:
  - Exchange rate volatility or perceptions about regime stability introduce uncertainty over the domestic price of imported investment goods, affecting investment responses after shocks.
  - Empirical and case evidence suggest the import content of investment goods in a subset of developing countries is about "30 percent" (Burstein et al., 2004).
  - Access to hedging instruments is often limited in developing countries; exchange rate uncertainty can therefore materially affect investment outcomes.
  - Even in some advanced-economy sectors, exchange rate uncertainty can negatively affect investment (Goldberg, 1993).

*Source: _wp0585 - Section V discussion.*

### 3.21 percentage point of GDP increase in the investment level (p-value=0.06). In contrast,

### _wp0585 - 3.21 percentage point of GDP increase in the investment level (p-value=0.06). In contrast,

### Investment: main findings
- Windstorms generate a cumulative increase in investment concentrated in developing countries with fixed exchange rate regimes.
- Estimated impacts (pooled and subgroup summaries from Table 3a and text):
  - A 3.21 percentage point of GDP increase in the investment level (p-value=0.06) is reported in the pooled three-year horizon context (text header).
  - In the “constant” regime subsample (exchange rate regime unchanged over a four year window beginning in the year before the shock), the estimated impact is a 1.91 percentage points of GDP increase in investment in developing countries with fixed exchange rate regimes.
  - The cumulative increase in investment over the three years remains around three percentage points of GDP in the basic results (text).
- Statistical significance and patterns from Table 3a (selected coefficients and standard errors preserved exactly as in source):
  - S_it * R_it in column (4) (DC): 0.0187** (0.009).
  - S_it+1 * R_it+1 in column (4) (DC): 0.013** (0.006).
  - S_it+1 * R_it+1 in column (6) (“Constant” Regimes (DC)): 0.019*** (0.007).
  - Note: *, **, *** indicates significance at the 10, 5 and 1 percent levels respectively.

### Endogeneity, regime choice, and robustness checks
- Potential selection bias and endogeneity:
  - Authors note a subtle selection issue: after a very severe windstorm a country may be unable to maintain a fixed rate regime and opt for increased flexibility, which could correlate flexible regimes with severe windstorms and bias estimates.
  - To address this, columns 5 and 6 of Table 3a restrict to “constant” regimes to reduce the possibility that the shock affected the exchange rate regime; results are modified only slightly, suggesting results are unlikely driven by shocks that systematically affected regime choice.
- Frequency and other country characteristics:
  - The frequency with which windstorms strike a country may influence regime choice and economic response; Column 2 of Table 3b allows the impact to depend on the percent of years with at least one windstorm over the last 100 years.
  - The basic results are robust when allowing the shock impact to vary by:
    - physical size (Table 3b column 3),
    - whether the country is an island (column 4),
    - per capita income (column 5),
    - the ratio of aid to gross national income (column 6),
    - and when all five variables are jointly included (column 7).

### Growth: main findings
- In the pooled sample (both advanced and developing countries) there is robust evidence that windstorms positively affect economic growth in the short run:
  - Growth is about 0.74 (p-value=0.07) percentage points higher in the year of the shock (S_t).
  - Growth is about 0.68 (p-value=0.02) percentage points higher in the year immediately afterwards (S_t+1).
  - No significant impact beyond the one year horizon in the pooled sample (text).
- Disaggregated by country type (Table 4a and text):
  - Advanced countries (Column 3):
    - Overall impact of the shock is relatively small.
    - Associated with a 0.53 percentage points of additional growth over the three years in economies with flexible versus fixed rate regimes (text).
    - Selected coefficients from Table 4a (AC column (3)):
      - S_t: -0.175 (0.188).
      - S_t+1: 0.726*** (0.254).
  - Developing countries (Column 4):
    - Cumulative impact over three years is economically large and slightly higher in fixed rate regimes:
      - 1.51 percentage points in fixed rate regimes versus 1.30 percentage points in flexible regimes (text).
    - Dynamics differ by regime:
      - Fixed rate economies: immediate declines in growth followed by large rebounds two years later.
      - Flexible regimes: large initial increases in growth, but no significant impact over the next two years.
    - Selected coefficients from Table 4a (DC column (4)):
      - S_t: 1.295** (0.602).
      - S_t * R_t: -1.473** (0.611).
      - S_t+1: 0.637 (0.425).
      - S_t+2 * R_t+2: 1.687*** (0.536).

### Model specification and controls
- Investment regressions (Table 3a / 3b) include linear controls: per capita income, the exchange rate regime, population density, percent of urban population, trade openness, and year and country fixed effects.
- Growth regressions (Table 4a) augment the core specification with: initial income per capita (1961), the ratio of investment to GDP, growth in government consumption (lagged one year), a measure of democracy, country size, growth in the terms of trade, Africa and Latin America indicator variables, ratio of exports and imports to GDP, urban population and population density, and year effects.
- Estimation details:
  - S_it is an indicator variable equal to one if a windstorm occurs in country i in year t and 0 otherwise.
  - R_it takes the value 1 if the exchange rate regime is fixed, and 0 if flexible.
  - “Constant” Regimes are those where the exchange rate regime remained constant over a four year window, beginning in the year before the shock.
  - Huber-White robust standard errors reported in parentheses; residuals clustered at the country level.

### Key statistics and sample sizes (selected)
- Table 3a NOBs (observations):
  - Pooled Sample: 2843
  - AC (column 3): 748
  - DC (column 4): 2095
  - “Constant” Regimes (AC) (column 5): 712
  - “Constant” Regimes (DC) (column 6): 2013
- Table 3a R-squared examples:
  - Pooled Sample: 0.55
  - AC: 0.67
  - DC: 0.58
- Table 3b NOBs:
  - Typically 2095 observations for most DC specifications; 2020 in columns involving foreign aid.
  - R-squared: 0.61 across reported DC specifications.
- Table 4a NOBs and R-squared:
  - Pooled Sample (column 2): NOBs 1954, R-squared 0.25
  - AC (column 3): NOBs 640, R-squared 0.44
  - DC (column 4): NOBs 1314, R-squared 0.27
  - “Constant” Regimes (AC) (column 5): NOBs 611, R-squared 0.44
  - “Constant” Regimes (DC) (column 6): NOBs 1268, R-squared 0.33

*Source: _wp0585 - 3.21 percentage point of GDP increase in the investment level (p-value=0.06). In contrast,*

### 0.43 percentage points less in fixed rate economies. The cumulative impact of windstorms in

### 0.43 percentage points less in fixed rate economies. The cumulative impact of windstorms in

### IV. ROBUSTNESS — overview
- This section performs further robustness analyses and probes mechanisms through which the choice of exchange rate regime affects adjustment after natural shocks.

### A. Regions: Africa, Asia, and the Caribbean
- Motivation:
  - Unobservable country characteristics and regional concentration of windstorms (Caribbean islands in Atlantic hurricane belt; Asian nations in Pacific typhoon belt) may confound inference.
  - Currency unions and long-standing pegs provide variation in exchange rate regime across regions (CFA franc zone, Eastern Caribbean Monetary Union, Hong Kong SAR pegs).
- Table 4b (developing countries robustness checks):
  - Cumulative impact of shock across regimes shows only minor variation when interacting shocks with:
    - frequency of windstorms (Column 2),
    - ratio of aid to gross national income (Column 3),
    - physical size (Column 4),
    - island status (Column 5),
    - all previous interaction terms jointly (Column 6).
  - Sharp differences in adjustment dynamics across regimes persist across these specifications.
- Regional findings from Table 5:
  - Sub-Saharan Africa:
    - One year after a windstorm, investment is about 5.30 percentage points higher in fixed rate regimes (Columns 2 and 3 of Table 5).
    - No significant impact on growth contemporaneously or one year later; two years after the shock growth is lower in flexible rate regimes (p-value=0.07). Footnote: That is, the hypothesis that Sit+2 + Sit+2 Rit+2 =0 is rejected at the 7 percent level.
  - Developing Asia:
    - No significant investment response detected.
    - Two years after the shock, growth is significantly higher in fixed rate regimes.
    - There is an indication that growth is lower in the year of the shock for fixed rate regimes, but not significant at conventional levels (p-value=0.11).
  - Caribbean:
    - Responses are similar to overall sample but with much larger point estimates, suggesting more dramatic impacts in small islands.
    - Data are missing for many Caribbean countries; growth specification uses a very parsimonious framework.

### B. Foreign Direct Investment (FDI)
- Rationale:
  - Replacing affected capital may attract FDI; FDI decisions are private-sector driven, helping isolate exchange rate regime effects on private investment behavior.
- Table 6 findings:
  - Pooled sample (advanced and developing countries, Column 2):
    - FDI response is positive and significant only for fixed rate regimes:
      - In the year of the shock (p-value=0.05).
      - One year after the shock (p-value=0.06).
  - Developing countries (disaggregated sample):
    - FDI response limited to the year of the shock and significant only in fixed rate regimes.
    - FDI increases by about 0.51 percentage points of GDP (p-value=0.10) in the year of the shock for developing countries in fixed regimes.
  - Qualitative conclusion: FDI response to natural shocks and the role of the exchange rate are qualitatively similar to gross capital formation.

### C. Earthquakes
- Motivation:
  - Earthquakes differ from windstorms: occur with almost no warning, not seasonal, and strike different country sets (islands are about five times more likely to experience windstorms than earthquakes).
  - Useful to test generalizability of results to different shock types and regions.
- Table 7 findings (earthquake shocks):
  - Investment:
    - Investment response to earthquakes is qualitatively similar to windstorms but with smaller point estimates than those for windstorms (Table 3a).
    - In developing countries with fixed rate regimes, investment is about 1.20 percentage points of GDP higher compared with flexible regimes.
    - Developing countries with flexible regimes incur a mean decline in investment of about 0.97 percentage points of GDP the year after an earthquake shock.
    - Advanced economies: no discernible impact on investment.
  - Growth:
    - Point estimates indicate a similar adjustment pattern as windstorms in developing countries but are not statistically significant.
    - Advanced economies show evidence that growth response is lower in fixed rate regimes—about 0.88 percentage points (p-value=0.02).
  - Interpretation:
    - Results using earthquakes are qualitatively similar to windstorm results.
    - Higher standard errors in growth response among developing countries reflect relative infrequency of earthquakes and higher volatility of output growth, making statistical detection harder.

### D. Different Time Periods
- Data note:
  - Incidence of windstorms in the database from 1980 through 2001 is almost double the number from 1960- (text truncated at source).  

*Italic: IMF working paper content unit _wp0585 — robustness, regional and shock-type analyses of exchange rate regime effects on investment, FDI, and growth after windstorms and earthquakes.*

### 79. Several factors might explain this difference, including environmental changes; the

### _wp0585 - 79. Several factors might explain this difference, including environmental changes; the

### Temporal variation in estimated impacts and data-reporting effects
- The database’s increased reporting of shocks is tested by estimating the base specification separately for 1960-80 and 1980-2001 for both investment and growth.
- Estimates from 1960-80 are generally small and in many cases not significant; results from 1980-2001 are qualitatively similar to earlier findings.
- Example finding: windstorms have no significant impact on investment levels in either developed or less developed countries in the pre-1980 subsample. Over 1980 through 2001:
  - the cumulative impact in developing countries with fixed rate regimes is 5.42 percentage points of GDP,
  - there remains no significant impact in developed countries, or in DCs with flexible regimes.
- Among developing countries, estimated relationships in the pre-1980 data are not significant.

### Table 7 (Earthquakes) — selected regression structure and notes
- Dependant variables covered include GDP Growth and the Ratio of Gross Capital Formation (Net of Inventories) to GDP.
- Regressions include the exchange rate regime, population density, percent urban population, trade openness, and year effects.
- Column 3 also includes per capita income and fixed effects. Column 2 includes initial per capita income (1961); one year lagged government consumption growth; terms of trade growth; democracy index; physical size; investment; Latin America and Africa indicator variables.
- Huber-White robust standard errors in parenthesis. Regression residuals clustered at the country level.
- Significance notation: *, **, *** indicates significance at the 10, 5 and 1 percent levels respectively.
- “AC” and “DC” include only developed and developing countries.

### Table 8 (Different Time Periods) — key coefficient patterns
- Separate estimations for 1960-1980 and 1981-2001 for GDP Growth and Gross Capital Formation (investment).
- Notable coefficient patterns (preserved exactly as reported in table excerpts):
  - For GDP Growth 1960-80 (AC): S_t = 0.525 (0.812); S_t+1 = 2.429*** (0.400); S_t+1*R_t+1 = -1.616*** (0.455).
  - For GDP Growth 1981-2001 (DC): S_t = 0.665 (1.408); S_t = -0.304 (0.207) appears in the table layout for other columns — see table for full matrix.
- Sample sizes (NOBs) reported by column, for example:
  - 1960-80 GDP Growth AC: NOBs = 243; 1981-2001 GDP Growth DC: NOBs = 409.
- R-squared values reported per column (preserved as in the table lines): e.g., 0.49, 0.10, 0.39, 0.31, 0.88, 0.71, 0.71, 0.74 (aligned to columns in table).

### Exchange rate classification (Section E) — finer and alternative classifications
- A “fine” 6-way classification (Reinhart & Rogoff) ranges from 1—no independent legal tender through 6—dual market with missing parallel market data; freely floating regimes are ranked 4; “freely falling” (ranked 5) remain excluded.
- Main findings:
  - Investment response in advanced economies remains uncorrelated with the choice of exchange rate regime.
  - Cumulative increase in investment for an economy classified as having no legal tender, pre-announced peg, or currency board arrangement is about 4 percentage of GDP, while the mean response in a freely floating regime is a percentage point of GDP decline in investment.
  - Growth is significantly larger in economies with fixed rate regimes two years after the shock among developing countries.
  - Growth rebounds faster in developed economies with flexible regimes.
- Alternative classification (Levy-Yeyati and Sturzenegger [LY-S] statistical approach, values 2 through 5):
  - LY-S performs relatively poorly in the investment specification but produces results broadly similar to the Reinhart and Rogoff system in the growth equation: growth is higher in developing economies with fixed rate regimes.

### Mechanisms (Section F) — channels linking exchange rate regime to adjustment
- Exchange rate uncertainty channel:
  - Use the standard deviation derived from monthly movements in the parallel exchange rate in the year before the shock (PSD_it-1) as a proxy for exchange rate uncertainty faced by the private sector.
  - PSD_it-1 is computed from monthly movements in the parallel rate the year before the shock; this avoids reflecting the effect of the current shock on the nominal exchange rate.
- Table 10 key findings (Parallel rate volatility, windstorms, and investment):
  - A one standard deviation increase in the parallel rate in the year before the shock is associated with a 2.25 percentage point decline in investment one year after the shock in developing countries (text statement).
  - Table 10 reported coefficients (preserved exactly as table entries):
    - S_it+1*PSD_it-1*10-04 for DC: -0.273** [0.142]
  - Observations: 645 (AC) and 1327 (DC).
  - R-squared: 0.72 (AC) and 0.59 (DC).
  - Note: S_it is an indicator variable taking value one if a windstorm occurs in country i in year t.
  - Huber-White robust standard errors in parenthesis/brackets; residuals clustered at country level.
  - Significance notation: *, **, *** indicates significance at the 10, 5 and 1 percent levels respectively.

### Export channel and timing of output responses (Table 11 discussion)
- Replaces trade openness measure (exports plus imports to GDP) with exports to GDP to isolate export-sector effects.
- Key pattern:
  - Column 2 indicates prior across-regime findings remain robust when using exports to GDP.
  - Column 3 allows the impact of the shock to vary by the ratio of exports to GDP; results differ significantly from Column 2, suggesting export behavior accounts for much of the difference between fixed and flexible regimes.
  - In the year of the shock, the impact of windstorms on output in flexible regimes depends on the export sector.
    - Although the two interaction terms Sit*Rit and Sit*Exports are individually insignificant, they are jointly significant at the five percent level (p-value=0.04), implying the increased output response observed in flexible regimes in the year of the shock may reflect an increase in exports in those economies.
  - No significant output response in later years for flexible regimes: S_{t+1}, S_{t+2}, and interaction terms with exports are individually and jointly insignificant.
  - Adjustment in fixed rate regimes: immediate decline in output followed by a rebound two years later.
- Column 4 of Table 11 estimates the export response to shocks; discussion indicates windstorms are associated with a cumulative (text cuts off in provided excerpt).

*Source: _wp0585 - 79. Several factors might explain this difference, including environmental changes; the*

### 9.45 percentage points of GDP (p-value=0.02) decline in exports over two years in fixed rate

### _wp0585 - 9.45 percentage points of GDP (p-value=0.02) decline in exports over two years in fixed rate

### Exchange rate regime, windstorms, and the export channel
- Empirical evidence shows a 9.45 percentage points of GDP (p-value=0.02) decline in exports over two years in fixed rate regimes.
- In developing countries with flexible exchange rate regimes, the year after a windstorm is associated with a 3.52 percentage point (p-value=0.05) depreciation in the real effective exchange rate.
- In regimes classified as fixed, the depreciation is 44 percent smaller (p-value=0.03).
- Using a larger sample and the parallel nominal exchange rate:
  - Parallel rate depreciates about 3.17 percentage points in flexible rate regimes one year after shock.
  - Fixed rate regimes experience only a 0.73 percentage point (p-value=0.09) depreciation.
- From the reported specifications, there is no significant impact on inflation one year after the shock.
- Synthesis: movements in the nominal exchange rate and their impact on exports provide an exchange rate channel that helps economies manage adverse real shocks; this channel explains a significant fraction of the difference in adjustment dynamics between fixed and floating regimes.

### Short-run GDP, investment and adjustment mechanisms (summary of Table 11 and related text)
- The analysis identifies two distinct short-run adjustment paths to windstorm shocks across regimes:
  - Flexible regimes: faster rebound in economic growth, partly driven by currency depreciation and higher exports.
  - Fixed regimes: negative growth in the first year after the shock, becoming significantly positive in the second year; adjustment characterized by higher investment in the first two years after shock rather than the exports channel.
- Key reported coefficients and statistics (as cited in the source text):
  - Export decline in fixed regimes: 9.45 percentage points of GDP (p-value=0.02) over two years.
  - Real effective exchange rate depreciation in flexible regimes one year after shock: 3.52 percentage point (p-value=0.05).
  - Depreciation reduction in fixed regimes relative to flexible: 44 percent smaller (p-value=0.03).
  - Parallel nominal exchange rate depreciation in flexible regimes: 3.17 percentage points one year after shock.
  - Parallel nominal exchange rate depreciation in fixed regimes: 0.73 percentage point (p-value=0.09).
  - No significant impact on CPI inflation reported (column 7).

### Long-run effects: investment and growth (summary of Table 12 and discussion)
- Long-run analysis uses average values over 1961-2001 and an indicator for economies that were predominantly fixed (indicator = 1 if fixed in more than 64 percent of years; 0 otherwise).
- Long-run findings:
  - Robust evidence (p-value=0.01) that windstorms positively affect the average investment level in economies with predominantly fixed exchange rate regimes.
  - In predominantly fixed regimes, a one standard deviation increase in the average number of windstorms is associated with a 2 percentage point increase in the average investment level over the period.
  - The mean frequency of windstorms has no significant impact on investment in economies that maintained a flexible regime for more than 36 percent of the years.
  - A one standard deviation increase in the average number of windstorms is associated with a 0.78 percentage point increase in average growth (p-value=0.06) only in those economies that predominantly maintained a fixed rate regime.
- Sample/threshold details reported:
  - The threshold for “mostly fixed” is 64 percent of years (the sample median).
  - Columns estimated separately for above/below the median yield similar results.

### Policy implications and interpretation
- Flexible exchange rate regimes:
  - Provide faster output adjustment to adverse real shocks through currency depreciation and export responses.
  - Benefit short-run macroeconomic adjustment speed.
- Fixed exchange rate regimes:
  - Tend to stimulate higher investment responses to shocks, particularly in the first two years after shock, which can translate into higher average investment and growth over multi-decade horizons in economies that predominantly maintain fixed regimes.
  - May forego the faster short-run adjustment benefits of depreciation but gain long-run productivity benefits from investment responses.
- Net interpretation:
  - There is merit to both views: flexible regimes facilitate faster short-run adjustment via exchange rate movements and the export channel; fixed regimes can encourage investment responses that yield long-run growth benefits.
  - Policy trade-off: faster adjustment under flexible regimes may come at the cost of forgoing longer-run productivity gains tied to the investment response seen under fixed regimes.

### Caveats, robustness and generalizability
- The study exploits randomness of natural disasters (windstorms) for identification, improving inference but raising generalizability questions:
  - It remains an open question whether “traditional” shocks (terms of trade, productivity) are sufficiently similar to natural shocks for the results to generalize.
  - Recent empirical work using terms of trade shocks also documents faster output response in flexible regimes, suggesting possible broader relevance.
- Potential selection and measurement concerns:
  - Averaging over long samples (1961-2001) raises the possibility of selection bias: mean shock frequency might influence regime choice.
  - The authors note results using “constant regimes” and controlling for shock frequency were not substantially different, suggesting limited selection bias.
  - Lack of data on storm intensity (wind speed, etc.) prevents ruling out the possibility that storm intensity affects regime choice; anecdotal evidence suggests frequency and intensity are correlated.

*Italicized source: _wp0585 - 9.45 percentage points of GDP (p-value=0.02) decline in exports over two years in fixed rate*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp0585.pdf_