## wpiea2025122-print-pdf

## Source details

**Canonical URL:** [wpiea2025122-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf.json)

---

### H3: Introduction — context, questions, and approach
- Remittances: important inflow for foreign exchange and household income in emerging and developing economies; for a handful of LAC countries remittances rose from less than 10 percent of GDP in 2000 to about 20 percent of GDP in 2023 (close to 30 percent of GDP for Nicaragua and Honduras).
- Policy concern: large remittance increases may cause real exchange rate (REER) appreciation, reducing tradable sector competitiveness.
- Research questions:
  - Can exchange rate policies mitigate REER appreciation pressures caused by remittance surges?
  - Do structural characteristics (e.g., import dependence) determine how remittances affect the REER?
- Empirical approach:
  - Data: large set of countries over 2008-21 (annual).
  - Estimator: dynamic panel with the Blundell-Bond estimator (system GMM).
  - Outcome: estimate causal effects of remittances (percent of GDP) on level of REER and compute impulse response functions for short- and long-run dynamics.

### H3: Key empirical findings (overview)
- Correlation analysis:
  - No correlation between remittances and REER overvaluation in fixed exchange rate regimes.
  - Positive correlation between remittances and REER overvaluation in flexible exchange rate regimes.
  - Potential endogeneity caveat noted.
- Causal (dynamic panel) results:
  - REER appreciates less and more slowly in fixed regimes than in flexible regimes after one year; no significant difference at shock onset nor in the long run.
- Role of macro structure:
  - High-import countries: higher remittances do not significantly increase demand for non-tradables → smaller REER appreciation.
  - High-remittance-receiving economies (remittances/GDP above global median): experience REER appreciation regardless of regime; appreciation more pronounced under flexible regimes.
  - Countries with relatively higher remittances and relatively lower imports: REER appreciates after a remittances shock regardless of regime.
- Robustness: results hold across alternative thresholds, alternative REER and reserves data (IMF IFS), alternative specifications, and an expanded sample (baseline includes 79 countries; alternative includes 146 countries).

### H3: Data and variable construction (key elements)
- Sample period: 2008 to 2021 (annual).
- REER (main): Darvas (2011, 2021) constructed series; robustness uses REER from IMF IFS.
- Remittances (percent of GDP): IMF Balance of Payment Statistics (BOP), including personal transfers and compensation of employees (BPM6 definition).
- GDP series: WEO (October 2024 vintage).
- Exchange rate regime classification: AREAER (2008–2021), using de jure and de facto categories; methodology revision date: February 2, 2009.
- Policy rates: BIS central bank policy rates dataset and IMF IFS; inflation from WDI; use averages of annualized monthly nominal interest rates.
- Financial openness: Chinn and Ito (2006) index (first principal component of four AREAER dummies).
- Foreign reserves per GDP: External Wealth of Nations Database (Lane and Milesi-Ferretti (2018)), updated to 2021; robustness uses reserves from BOP and IIP (IMF).
- Controls: GDP per capita, current account, terms of trade, imports, tradable output, government expenditures (from WDI). Remittances excluded from current account in analysis.

### H3: REER overvaluation measure (section 4)
- Long-run benchmark: estimate log REER_it = α + β log GDPPC_it + δ_t + ε_it; overvaluation = ε̂_it = log REER_it − log REER_it^.
- Interpretation: residual captures deviations of REER from long-run component predicted by GDP per capita and time effects (Balassa-Samuelson logic preserved).
- Empirical pattern (2008–2021):
  - Fitted lines by regime: overall fitted line shows no strong positive correlation between remittances and overvaluation; fitted line for floating arrangements shows positive correlation (prima facie Dutch disease in floating regimes).
  - Causality qualification: endogeneity may bias correlation (e.g., overvaluation could lower remittances).

### H3: Empirical strategy and dynamic panel specification
- Objective: estimate remittance impact on REER and heterogeneity by exchange rate regime; compute impulse responses up to 4 years ahead.
- Model (estimated separately by regime):
  - log REER_it = α + Σ_{s=1}^{l} β_s log REER_{it−s} + Σ_{s=0}^{m} γ_s (Remittance_{it−s} / GDP_{it−s}) + X'_it δ + ν_i + φ_t + ε_it
  - Controls X_it: log real GDP per capita; terms of trade; current account; real interest rate differential relative to the U.S.; financial openness; foreign reserves (percent of GDP); government expenditure (percent of GDP).
- Estimator: Blundell and Bond (system GMM) to address endogeneity; use levels of REER as dependent variable to assess impact at shock onset, after 1 year, and long term.

### H3: Baseline estimation results (section 6)
- Impulse responses (fixed vs flexible regimes):
  - Fixed regimes:
    - REER begins to appreciate one year after remittance shock, peaks in two years.
    - Sluggish appreciation attributed to short-run price stickiness; no statistically significant effects over 4-year horizon.
  - Flexible regimes:
    - REER appreciates more within one year than under fixed regimes; immediate and stronger response likely due to nominal exchange rate appreciation.
    - Wider confidence intervals due to nominal exchange rate volatility.
    - No statistically significant initial appreciation at shock onset, but effect after one year is larger and statistically significant.
    - No lasting long-run change; initial deviations unwind.
  - Overall: appreciation begins after one year under both regimes; stronger under flexible regimes; no statistically significant long-run effect in baseline estimations.
- Subsample results:
  - Importers:
    - Low importers under fixed regimes: greater REER appreciation than high importers.
    - High importers: no statistically significant effect.
    - Flexible regimes: higher and quicker appreciation.
  - Size of economy:
    - Split by GDP median: REER appreciates more in large countries.
    - Both large and small economies show statistically significant appreciation, especially in floating arrangements.
  - High remittance receivers (baseline shock = remittance-to-GDP increases by 1 percentage point):
    - High remittance receivers: statistically significant REER appreciation under fixed regimes.
    - High receivers: appreciation after remittance increases regardless of regime; flexible regimes show higher and quicker appreciation.
    - Low remittance receivers: significant appreciation under flexible regimes.
  - Remittances and imports jointly:
    - High remittances & low imports: REER appreciates after remittance shock regardless of regime.
    - Low remittances & high imports: highest peak REER appreciation (estimates less precise; subgroup has 45 observations and 8 countries).
    - High remittances & high imports & flexible regimes: no significant effect; wide confidence intervals.
  - Remittances, imports, and REER undervaluation:
    - Remittance shock impact higher under floating arrangements; appreciation may be larger when REER is undervalued.
    - Fixed regimes: no statistically significant impact regardless of undervaluation/overvaluation.
    - Low remittances & undervaluation: highest impact on REER appreciation.

### H3: Robustness analysis (section 7) — key checks and quantitative notes
- Different thresholds for imports (70th/30th percentiles):
  - Main estimation: 149 observations; robustness check: 77 observations.
  - After one year, REER does not appreciate at 95 percent confidence; low importers under flexible regime: REER appreciates significantly at 90 percent level (p-value = 0.054).
  - Confirms baseline direction: low importers experience more appreciation; wider CI due to fewer observations.
- Different thresholds for remittances (80th/20th percentiles):
  - Subgroup size smaller (50 observations vs 149).
  - Fixed regimes: slower and smaller REER appreciation, consistent with main estimation.
  - No statistically significant appreciation for high remittance receivers under flexible regime in this check (wide CIs).
- REER from IFS (smaller country coverage):
  - Sample: 172 observations and 19 countries (versus main fixed-regime regression: 385 observations with 43 countries).
  - Flexible regime: no significant appreciation at 95 percent, significant at 90 percent (p-value = 0.083).
  - Fixed regime: smaller effect; REER appreciates one year after shock.
- Foreign reserves from BOP and IIP (fewer observations):
  - Right panel: no significant appreciation at 95 percent, significant at 90 percent (p-value = 0.105).
  - Left panel: smaller REER appreciation after shock; consistent direction with baseline.
- Larger sample excluding policy rates (to include countries missing policy rate data):
  - New sample: 146 countries; period 2008-21.
  - Results: floating regimes experience sharp appreciation in year 1; fixed regimes show gradual appreciation—consistent with baseline.
  - Overidentification tests: instruments valid after excluding policy rates.
- Robustness summary:
  - Low importers generally show more REER appreciation after remittance shocks (significance sensitive to sample size).
  - Fixed regimes exhibit smaller and slower appreciation relative to flexible regimes across checks.
  - Evidence of REER appreciation under flexible regimes often significant at 90 percent even when not at 95 percent (reported p-values include 0.054, 0.083, 0.105).
  - Loss of statistical significance in some checks largely attributed to smaller sample sizes (examples: 77 vs 149; 50 vs 149; 172 vs 385 observations).

*Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Context and motivation
- Remittances are an important inflow for many emerging and developing economies for foreign exchange and household income support.
- For a handful of LAC countries, remittances grew from less than 10 percent of GDP in 2000 to about 20 percent of GDP in 2023 (close to 30 percent of GDP for Nicaragua and Honduras), well above the world’s and LAC averages.
- Policymakers are concerned that large increases in remittance flows may cause real exchange rate (REER) appreciation, reducing competitiveness of the tradable goods sector.

### Research questions and approach
- Two key dimensions studied:
  - Whether exchange rate policies can mitigate appreciation pressures on the REER caused by surges in remittance flows.
  - Whether structural country characteristics—such as import dependence—determine how remittance flows affect the REER.
- Empirical strategy:
  - Use data from a large set of countries over 2008-21.
  - Employ a dynamic panel model with the Blundell-Bond estimator to estimate causal effects of remittances (as percent of GDP) on the level of the REER while controlling for numerous determinants of exchange rate behavior.
  - Compute impulse response functions to analyze short- and long-run dynamics.

### Key empirical findings
- Correlation analysis (following Rajan and Subramanian (2005) and Rodrik (2008)):
  - Remittance flows are not correlated with REER overvaluation in economies operating under fixed exchange rate regimes.
  - Remittance flows are positively correlated with REER overvaluation in economies with flexible exchange rate regimes.
  - Note: potential endogeneity may influence the correlation result.
- Causal estimation (dynamic panel with Blundell-Bond):
  - The REER appreciates less and more slowly in countries with fixed exchange rate regimes than in those with flexible regimes after one year, but not at the onset of a remittances shock, nor in the long run.
- Role of macroeconomic structure:
  - In countries with a large share of imported goods, higher remittances do not significantly increase demand for non-tradable goods, resulting in smaller REER appreciation.
  - High-remittance-receiving economies—defined as those receiving remittances as share of GDP above the global median—experience REER appreciation regardless of exchange rate regime; appreciation is more pronounced in countries with flexible exchange rate regimes.
  - Countries with relatively higher remittances and relatively lower imports experience REER appreciation after a remittances shock, regardless of exchange rate regime.
- Robustness:
  - Results are robust to alternative classifications by imports and remittance flows using different thresholds, alternative data sources for REER and foreign reserves (IMF IFS), alternative specifications, and an expanded sample (baseline includes 79 countries; alternative includes 146 countries).

### Data and key variables (overview)
- Sample: annual observations from 2008 to 2021 (primary analysis).
- REER data: Darvas (2011, 2021) constructed REER series (main regressions); robustness checks use REER from the IMF’s International Financial Statistics (IFS).
- Remittances (percent of GDP): IMF Balance of Payment Statistics (BOP); remittances include personal transfers and compensation of employees as defined by the sixth edition of the IMF's Balance of Payments Manual. GDP series from WEO (October 2024 vintage).
- Exchange rate regime classification: IMF Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER), using both de jure and de facto categories; AREAER covers 2008 to 2021 classifications. Methodology revision date: February 2, 2009.
- Policy rates and real interest rates: policy rates from BIS central bank policy rates dataset and IMF IFS; inflation from World Development Indicators (WDI). Use averages of annualized monthly nominal interest rates for each year.
- Financial openness: Chinn and Ito (2006) index (first principal component of four AREAER dummies).
- Foreign reserves per GDP: External Wealth of Nations Database (Lane and Milesi-Ferretti (2018)), updated to 2021; robustness uses foreign reserves from BOP and IIP (IMF).
- Controls: GDP per capita, current account, terms of trade, imports, tradable output, government expenditures (from WDI). Remittances are excluded from the current account in the analysis.
- Country classification for imports and tradable goods producers uses imports and tradable output.

### Contribution to the literature
- Simultaneously addresses:
  - How the effect of remittance flows varies across exchange rate regimes.
  - Short- and long-run effects of remittance flows on the REER via impulse response functions.
  - Inclusion of control variables relevant to exchange rate determination.
  - How macroeconomic structural characteristics (levels of remittances and imports) influence remittance effects on the REER.

### Paper structure (as presented)
- Section 2: Data
- Section 3: Related Literature
- Section 4: Remittance and Overvaluation of the Real Effective Exchange Rate
- Section 5: Empirical Strategy
- Section 6: Baseline Estimation Results
- Section 7: Robustness Analysis
- Section 8: Conclusion

*Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf*

### 4.  Remittance and Overvaluation of the Real

### 4.  Remittance and Overvaluation of the Real

### Effective Exchange Rate
- Purpose: construct a measure of real effective exchange rate (REER) overvaluation that captures the deviation of the REER from its long-term component (GDP per capita as predicted by the Balassa-Samuelson model).
- The Balassa-Samuelson model logic preserved:
  - Richer countries with higher productivity in tradable goods sector have a higher price level due to higher prices of non-tradable goods.
  - Productivity increases in tradables → higher wages in tradables → higher prices for non-tradables.
- Empirical implementation:
  - Estimate: log REER_it = α + β log GDPPC_it + δ_t + ε_it
    - REER_it is the real effective exchange rate of country i in year t.
    - GDPPC_it is GDP per capita for country i in year t.
    - δ_t are time-fixed effects.
    - ε_it is the error term.
  - Predicted log REER: log REER_it^ = α̂ + β̂ log GDPPC_it + δ̂_t.
  - Overvaluation measure: residual ε̂_it = log REER_it − log REER_it^.
    - Interpretation: residual captures components of REER unrelated to GDP per capita or time trend after excluding long-run component and time-fixed effects.
- Notes on long-run benchmark:
  - The long-run REER benchmark is assumed to be time-invariant, though structural changes (e.g., trade liberalization, supply shocks) may shift equilibrium REERs over time.
- Empirical relationship (Figure 3, 2008 to 2021):
  - Economies classified by exchange rate regimes in 2021 (robust to classification by most frequent regime during sample).
  - Exchange rate regime classification (AREAER broad measure):
    - Hard pegs: “Exchange arrangement with no separate legal tender”, “Currency board arrangement.”
    - Soft pegs: “Conventional pegged arrangement”, “Stabilized arrangement”, “Crawling peg”, “Crawl-like arrangement”, “Pegged exchange rate within horizontal bands.”
    - Floating arrangements: “floating”, “free-floating.”
  - Fitted lines:
    - Black line: fitted line for all countries — does not show a strong positive correlation between remittances and the overvaluation measure.
    - Red line: fitted line for countries under floating arrangements — remittances are positively correlated with the overvaluation measure.
  - Interpretation: prima facie evidence of a Dutch disease effect in countries under floating arrangements; but causality requires addressing endogeneity (e.g., overvaluation may reduce remittances by lowering value of unit remittance, potentially underestimating appreciation pressure of remittances).

### Empirical Strategy
- Objective: estimate the effect of remittances on the REER and evaluate heterogeneity by exchange rate regime.
- Time horizon: capture both short- and long-run effects; use impulse response functions of a remittance shock via a dynamic panel model.
- Omitted variables concern: GDP per capita proxies productivity in tradables but may correlate with terms of trade, aid dependence, financial openness; thus specification also controls for these variables.
- Subsample approach:
  - Separate economies into fixed exchange rate regime and flexible exchange rate regime.
  - Fixed regime subsample includes both hard pegs and soft pegs (limited number of hard pegs in sample).
- Frequency: annual observations (some variables unavailable at higher frequency: terms of trade, financial openness, government expenditure).
- Endogeneity and estimator:
  - Use the Blundell and Bond estimator (system GMM variant): first differences of lagged dependent variables used as instruments under assumption that error terms are not correlated with these instruments.
- Choice of dependent variable:
  - Use level of the REER (not change) to align with motivation from Figure 3 and to assess impact of remittance levels on REER at shock onset, after a year, and in the long-term.
  - Impulse responses shown until 4 years ahead.

### Dynamic Panel Model (Specification)
- Model estimated for each subsample (floating and fixed regimes):
  - log REER_it = α + Σ_{s=1}^{l} β_s log REER_{it−s} + Σ_{s=0}^{m} γ_s (Remittance_{it−s} / GDP_{it−s}) + X'_it δ + ν_i + φ_t + ε_it
    - Lags of REER included to capture dynamics and error-correction tendencies.
    - Lags of remittances (as percent of GDP) included to compute impulse response functions; coefficients on lagged remittance-GDP ratio capture dynamic effects of remittances on REER.
    - ν_i denotes country fixed effects; φ_t denotes time fixed effects.
- Controls included (X_it):
  - log of real GDP per capita
  - terms of trade
  - current account
  - real interest rate differential relative to the U.S.
  - financial openness
  - foreign reserves (as percent of GDP)
  - government expenditure (as percent of GDP)
- Rationale for controls:
  - GDP per capita: key long-run determinant of REER.
  - Current account and foreign reserves: affect REER.
  - Terms of trade: determine demand for non-tradable goods, influencing REER.
  - Policy-related variables (central bank policy rates proxied by real interest rate differential, financial openness, foreign reserves, government expenditure) affect real exchange rates.
  - Time fixed effects capture global factors such as GDP in the U.S. or VIX.
- Impulse response interpretation:
  - Annual frequency → impulse responses capture impact of remittance levels on REER at shock onset, after 1 year, and up to 4 years ahead.

*Source: IMF Working Papers — chapter "4.  Remittance and Overvaluation of the Real" (extracted content).*

### 6.  Baseline Estimation Results

### 6.  Baseline Estimation Results

### A. Impulse Response Functions Estimated by Dynamic Panel Model
- Estimated impulse response functions of the real effective exchange rate (REER) to a one-off increase in remittance flows using a dynamic panel model on two subsamples: fixed exchange rate (hard pegs and soft pegs) and flexible exchange rate regimes.
- Annual-frequency data: impulse responses illustrate impact in subsequent years.
- Fixed exchange rate regime:
  - REER begins to appreciate one year after the remittance shock, reaching peak appreciation in two years.
  - Sluggish appreciation attributed solely to changes in price levels; price stickiness in short run explains lack of immediate REER appreciation.
  - No statistically significant effects of the remittance shock over the 4-year post-shock horizon.
- Flexible exchange rate regime:
  - REER appreciates more within one year than under fixed regimes; immediate and stronger response likely due to nominal exchange rate appreciation following the remittance shock.
  - Wider confidence intervals across years due to nominal exchange rate volatility.
  - No statistically significant initial appreciation at the onset of the shock (price stickiness possible), but the effect after one year is larger and statistically significant.
  - In the long run, no lasting change in the REER—initial deviations are fully unwound.
- Overall summary:
  - Under both regimes, REER begins to appreciate one year after the remittance shock.
  - Appreciation is stronger under flexible regimes due to nominal exchange rate fluctuations; under fixed regimes, nominal rate is stabilized and short-term prices are sticky.
  - No statistically significant long-run effect of remittances on REER in baseline estimations.
  - Analysis estimates impact on level of REER, not deviation from equilibrium; steady increases in remittances over time could contribute to overvaluation in fixed-rate countries.

### B. Importers
- Hypothesis: In countries that import more foreign goods, remittances are more likely spent on imported tradable goods, reducing pressure on demand for non-tradables and thus muting REER appreciation.
- Method: Define high (low) importers by median import-to-GDP ratio relative to global median; split by exchange rate regime; estimate IRFs for four subgroups.
- Findings:
  - Low importers under fixed regimes experience greater REER appreciation following a remittance shock than high importers under the same regime.
  - No statistically significant effect for high importers; low importers show REER appreciation due to remittance shocks.
  - Flexible regimes display higher and quicker REER appreciation compared to fixed regimes, supporting baseline finding.

### C. Size of the Economy
- Observation: Smaller open economies have higher import shares in consumption; countries that import fewer goods tend to be larger economies.
- Method: Split sample into small and large countries by GDP (below or above the median).
- Findings:
  - REER appreciates more in large countries.
  - Both large and small economies experience statistically significant REER appreciation, especially in floating exchange rate arrangements.

### D. High Remittance Receivers
- Classification: High (low) remittance receivers receive remittances above (below) the world median of remittances relative to GDP.
- Baseline shock considered: remittance-to-GDP ratio increases by 1 percentage point.
- Findings:
  - High remittance receivers experience statistically significant REER appreciation under fixed exchange rate regimes.
  - For high remittance receivers, REER appreciates following remittance increases regardless of exchange rate regime.
  - As with other results, flexible regimes exhibit higher and quicker REER appreciation.
  - Low remittance receivers show significant REER appreciation under flexible exchange rate regimes; magnitude differences across panels are not large.

### E. Remittances and Imports Jointly
- Method: Classify countries by (i) level of imports, (ii) level of remittances, and (iii) exchange rate regime; estimate IRFs.
- Findings:
  - Countries with high remittances and low imports: REER appreciates after a remittance shock regardless of exchange rate regime.
  - Countries with low remittances and high imports have the highest REER appreciation at the peak—results possibly driven by nominal exchange rate movements.
  - Subsample with low remittances & high imports is the smallest: 45 observations and 8 countries (about 6 observations per country on average), making estimates less precise.
  - Countries with high remittance, high imports, and flexible regimes show no significant effect and have wide confidence intervals due to nominal exchange rate volatility.

### F. Remittances, Imports and REER Undervaluation
- Hypothesis: Effect of remittance shock on REER depends on degree of REER undervaluation/overvaluation and the extent remittances are spent on imports versus domestic goods.
- Method: Split sample between countries with overvalued REER and countries with undervalued REER (using estimations from section 4).
- Findings:
  - Impact of a remittance shock on REER appreciation is higher under floating arrangements, as nominal exchange rate adjusts to incoming foreign exchange.
  - Appreciation may be more significant in cases of undervaluation because there is more room for REER to appreciate before reaching equilibrium.
  - In fixed exchange rate regimes, impact on REER is not statistically significant regardless of departure from equilibrium.
  - Countries with low remittances and undervaluation show the highest impact of a remittance shock on REER appreciation.

*IMF Working Paper — Chapter 6: Baseline Estimation Results (impulse response functions of REER to remittance shocks across regimes and country characteristics).*

### 7.  Robustness Analysis

### 7.  Robustness Analysis

### A. Different Thresholds for Imports
- Method:
  - High importers: imports (percent of GDP) above the 70th percentile.
  - Low importers: imports (percent of GDP) below the 30th percentile.
  - Chosen 70th/30th percentiles to ensure enough observations for IRF estimation.
- Findings:
  - Figure 11: IRFs of the REER by (Non-)Importers and exchange rate regime.
  - After one year of the remittance flow shock, the REER does not appreciate significantly at the 95 percent confidence level.
  - REER appreciates significantly at the 90 percent level and p-value is 0.054 for low importers under flexible exchange rate regime.
  - Confirms baseline: low importers experience more appreciation.
  - Unlike main estimation, no statistically significant appreciation for low importers adopting fixed exchange rate regime in this robustness check.
- Interpretation:
  - Wider confidence intervals likely due to fewer observations: main estimation has 149 observations; this robustness estimation has 77 observations, causing wider confidence intervals.

### B. Different Thresholds for Remittances
- Method:
  - High remittance receivers: remittance flows above the 80th percentile.
  - Low remittance receivers: remittance flows below the 20th percentile.
- Findings:
  - Figure 12: IRFs by high (low) remittance receivers and exchange rate regimes with different threshold.
  - Countries under fixed exchange rate regime experience slower and smaller appreciation of the REER, consistent with main estimation.
  - No statistically significant appreciation of the REER for high remittance receivers under flexible exchange rate regime in this robustness check.
- Interpretation:
  - Wide confidence intervals for the subgroup likely because only 50 observations are available here versus 149 observations in the main subgroup estimation.

### C. Real Effective Exchange Rates from IFS
- Method:
  - Replace REER constructed by Darvas (2011, 2021) with REER data from the IFS of the IMF.
  - Note: IFS data covers a smaller number of countries.
- Findings:
  - Figure 13: Impulse Response Functions of the REER from IFS of the IMF.
  - Flexible regime (right panel): REER does not appreciate significantly at the 95 percent confidence level, but appreciates significantly at the 90 percent confidence level (p-value is 0.083), consistent with baseline.
  - Fixed regime (left panel): effect of remittance flows on the REER is smaller, but the REER appreciates one year after the shock.
- Sample sizes:
  - This analysis: number of observations is 172 and number of countries is 19.
  - Main regression for the fixed exchange rate regime: 385 observations with 43 countries.
- Interpretation:
  - Differences could be due to fewer observations in the IFS-based REER analysis.

### D. Foreign Reserves from IFS (BOP and IIP)
- Method:
  - Use foreign reserves data from the BOP and IIP of the IMF (Lane and Milesi-Ferretti (2018) extended dataset from BOP and IIP).
  - Note: these data have fewer observations.
- Findings:
  - Figure 14: Impulse Response Functions of the REER (Foreign Reserves from BOP and IIP of the IMF).
  - Right panel: does not show statistically significant appreciation of the REER after the remittance shock at the 95 percent confidence level, but significant at 90 percent confidence level (p-value is 0.105).
  - Left panel: REER appreciation is smaller after the shock of remittances.
  - Consistent with baseline: REER appreciates more under the fixed exchange rate regime, although it appreciates following the remittance shock.
- Interpretation:
  - Reduced significance and smaller effects may be due to fewer observations in the reserves dataset.

### E. Larger Sample (excluding policy rates as control)
- Method:
  - Exclude policy rates to include countries missing policy rate data (e.g., Haiti, Jamaica, Nicaragua, Honduras).
  - New sample includes 146 countries; sample period remains 2008-21.
- Findings:
  - Figure 15: IRFs of the REER (146 countries, no policy rate as control).
  - Baseline results robust to larger sample:
    - Countries with floating exchange rate regimes experience sharp appreciation in year 1.
    - Those with fixed exchange rate regimes exhibit more gradual appreciation.
- Instrument validity:
  - Overidentification tests implemented to assess instrument validity after excluding policy rates.
  - The overidentification test in both regressions imply that these instruments are valid.

### Key robustness patterns across checks
- Across alternative thresholds, data sources, and sample sizes:
  - Low importers generally exhibit more REER appreciation after remittance shocks (significance varies with sample size).
  - Fixed exchange rate regimes exhibit smaller and slower REER appreciation relative to flexible regimes in multiple robustness checks.
  - Evidence of REER appreciation under flexible regimes is often significant at the 90 percent confidence level even when not at 95 percent, with reported p-values including 0.054, 0.083, and 0.105 in different checks.
- Data- and sample-size sensitivity:
  - Many loss of statistical significance attributed to smaller sample sizes in robustness exercises (examples: 77 observations vs 149; 50 observations vs 149; 172 observations vs 385).

*Source: wpiea2025122-print-pdf - 7.  Robustness Analysis*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025122-print-pdf.pdf_