## wpiea2024244-print-pdf - Introduction

## Source details

**Canonical URL:** [wpiea2024244-print-pdf - Introduction](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024244-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024244-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024244-print-pdf.pdf.json)

---

### Introduction and scope
- Document examines links between uncertainty and remittances.
- Main thematic areas:
  - "Uncertainty and Remittances: How Close are the Links?"
  - "Empirical Strategy"
  - "Remittance and Uncertainty Data and their Measurements"
  - "The Model, Sample and Econometric Approach"
  - "Empirical Findings"
  - "Conclusion"
- Empirical strategy sections and annexes include:
  - Empirical Strategy; Remittance and Uncertainty Data and their Measurements; The Model, Sample and Econometric Approach.
  - Annexes: Country Sample Composition; Summary statistics and correlation matrix; Correlation matrix; Variable Definitions and Sources; Stationarity tests: Maddala and Wu (1999).

### Data, measurements, and sample
- Uncertainty index:
  - Source: Ahir, Bloom and Furceri (2022); constructed from EIU quarterly country reports (mentions of “uncertainty” normalized by total word count).
  - Coverage in source database: 143 countries, 1952Q1–2021Q2 (figures use developing-country series 1995Q1–2020Q1).
- Remittances dataset:
  - Novel quarterly dataset for 95 countries (18 high-income, 62 middle-income, 15 low-income) building on Kpodar et al. (2023).
  - Coverage: from 1971Q1 (a few countries) through 2020Q4 for most countries; consolidated at receiving-country level.
  - Definition: personal transfers and compensation of employees; proxies used when components missing.
- Final empirical sample: 77 developing countries, 1999Q1─2019Q4.
- Key controls: domestic and foreign GDP per capita, total migrant stock, domestic currency per U.S. dollar (level and volatility), exchange rate volatility (quarterly standard deviation).

### Empirical strategy and baseline model
- Baseline linear specification (Eq(1)):
  - log(Rem)_{c,t} = γ + α Uncert_{c,t}^l + β Uncert_{c,t}^f + Θ X_{c,t} + μ_c + ϑ_t + ε_{c,t}
  - Rem_{c,t}: remittances in millions of US dollars.
  - Uncert^l: domestic (local) economic uncertainty.
  - Uncert^f: foreign (host-country) economic uncertainty (migrant-weighted average across host countries).
  - X: controls (income per capita of receiving and sending countries, migrant stock, exchange rate level and volatility).
  - Country fixed effects and time dummies included; robust standard errors.
- Dynamic analysis:
  - Local projection (LP) specification (Eq(2)) with n = 4 lags (quarterly data) and horizon h up to 5 quarters.
  - Includes Teulings and Zubanov (2014) correction to control for subsequent shocks.
- Endogeneity approach:
  - Instrumental variables: presidential/legislative election dummies and export-weighted partner uncertainty.

### Main empirical findings — baseline and fixed-effect estimations
- Domestic uncertainty:
  - Baseline: domestic uncertainty does not significantly influence remittance magnitudes across multiple specifications (columns 1, 3, 4, and 6 in Table 1).
  - Heterogeneous effects: net effect can be positive or negative depending on country characteristics—domestic uncertainty increases remittances where investment motives are weak and public provision of health and education is poor; it reduces remittances where remittances are channeled to investment.
- Foreign (host-country) uncertainty:
  - Robust negative effect on remittances across specifications.
  - Representative coefficients (Table 1): -0.238 [0.102]**, -0.223 [0.084]***, -0.225 [0.076]***, -0.225 [0.086]** (columns 2, 3, 5, 6).
- Selected control-variable coefficients (Table 1 examples):
  - GDP per capita (domestic): 1.271 [0.321]***, 1.090 [0.301]***, 1.260 [0.320]***, 1.250 [0.305]***.
  - Foreign GDP per capita (log): 0.390 [0.093]***, 0.449 [0.083]***, 0.385 [0.093]***.
  - Domestic currency per U.S. dollar (log): 0.237 [0.109]**, 0.184 [0.111], 0.242 [0.110]**.
  - Exchange rate volatility: -0.384 [0.209]*, -0.418 [0.240]*, -0.387 [0.207]*.

### Robustness checks and alternative specifications
- Alternative remittance measures (Table 2): domestic uncertainty non-significant; foreign uncertainty negative (example: -0.424 [0.217]* for remittances per migrant).
- Additional controls (Table 3): financial access, inflation, monetary policy rate, GDP growth, remittance transaction costs do not alter core results; foreign uncertainty remains negative and significant (examples: -0.209 [0.092]**, -0.224 [0.086]**, -0.172 [0.070]**).
- Instrumental variables (Table 4): instrumenting domestic uncertainty yields exogenous component statistically insignificant for remittances while foreign uncertainty stays negative (example: -0.229 [0.092]**).
- Results robust to growth-volatility measures (standard deviation of annual real GDP growth over preceding 5 years), inclusion of remittance costs, and alternative remittance indicators (levels, per capita, ratios).

### Heterogeneity: altruism versus investment channels
- Investment channel:
  - Higher private investment-to-GDP ratio → larger share of remittances for investment → domestic uncertainty reduces remittances via lower expected returns.
  - Evidence (Table 5): lagged private investment-to-GDP ratio positive and significant: 0.012 [0.004]***, 0.016 [0.004]***; interaction Inv*Domestic uncertainty negative and significant: -0.025 [0.010]**, -0.021 [0.009]**.
- Altruism / public-service channel:
  - Low public spending on health and education → domestic uncertainty elicits higher remittances for consumption-smoothing.
  - Evidence (Table 5): interaction terms Health*Domestic uncertainty and Educ*Domestic uncertainty negative and significant in relevant columns (e.g., Health*Domestic uncertainty [0.000]**, Educ*Domestic uncertainty [0.003]**), implying stronger positive remittance responses where public provision is weak.
- Thresholds and marginal effects:
  - LP-based threshold analysis (Figure 8): marginal impact of domestic uncertainty on remittances crosses zero at private investment ratio = 18 percent of GDP.
  - Sample median private investment ratio = 15.14 percent; sample average = 16 percent.
  - Example high private investment observation: 90th percentile Private Investment Ratio to GDP = 25.3 percent.

### Dynamics — Local projection (LP) results
- Domestic uncertainty:
  - LP shows no discernible contemporaneous correlation with remittances once dynamics are accounted for (Figure 6).
- Foreign uncertainty:
  - LP shows significant negative effect on remittances in the initial quarter following a foreign uncertainty shock; effect attenuates and becomes statistically insignificant thereafter (Figure 7).
- LP heterogeneity (private investment percentiles):
  - Low private investment (10th percentile): domestic uncertainty associated with higher remittances in the first quarter only; effects dissipate after quarter 1.
  - High private investment (90th percentile): domestic uncertainty yields a negative and significant effect in quarter 1.
  - Marginal impact crosses zero at twenty (18) percent of GDP private investment ratio.

### Responses conditional on public spending (health and education)
- Public health spending per capita:
  - Low public health spending associated with a positive remittance response to a rise in uncertainty; high public health spending associated with a delayed negative response.
  - Example percentiles from Figure 9: 10th percentile Public Health Spending per Capita = US$22.5; 90th percentile = US$524.0.
  - Timing: marginal impact becomes significant only beyond the first quarter; in high-public-health-spending countries remittances decline only after the fourth quarter; in low-public-health-spending countries positive response occurs from the second quarter.
- Public education spending per capita:
  - Similar pattern to health: 10th percentile Public Education Spending per Capita = US$1.1; 90th percentile = US$35.4.
  - Remittances increase with uncertainty where public education investment is low; remittances decline with a lag where public education investment is high.
- Confidence intervals: shaded areas in figures represent the 90 percent confidence interval.

### Annex 2 key statistics (selected rows; exact values preserved)
- Sample: 77 countries (1999Q1 2019Q4).
- Remittances (millions of US dollars, log): Obs 4,188; Mean 5.408; Std. dev. 2.042; Min -1.738; Max 10.030
- Remittances per migrant (thousands of US dollars, log): Obs 4,188; Mean -1.256; Std. dev. 1.548; Min -7.412; Max 8.226
- Remittances to GDP (log): Obs 4,188; Mean -0.299; Std. dev. 1.634; Min -6.695; Max 2.428
- Domestic uncertainty index: Obs 3,580; Mean 0.200; Std. dev. 0.215; Min 0.000; Max 2.038
- Foreign uncertainty index: Obs 4,188; Mean 0.219; Std. dev. 0.147; Min 0.001; Max 1.668
- Private investment ratio to GDP: Obs 3,538; Mean 17.838; Std. dev. 6.880; Min -7.019; Max 54.799
- Public spending on heath per capita (US dollars): Obs 3,926; Mean 223.272; Std. dev. 202.781; Min 8.362; Max 1192.82
- Public spending on education per capita (US dollars): Obs 2,942; Mean 16.084; Std. dev. 15.230; Min 0.479; Max 86.432
- Public spending on health and education per capita (US dollars): Obs 2,856; Mean 241.890; Std. dev. 223.231; Min 9.264; Max 1242.15

### Stationarity (Annex 5 — Maddala and Wu (1999) results; exact reported values)
- Remittances (log): Level Chi2 Stat.: 348.627; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Remittances per migrant (log): Level Chi2 Stat.: 352.677; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Remittances to GDP (log): Level Chi2 Stat.: 550.842; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Remittances per capita (log): Level Chi2 Stat.: 420.283; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Remittances per migrant as a share of income per capita of the host economy (log): Level Chi2 Stat.: 444.437; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Domestic uncertainty index: Level Chi2 Stat.: 2,951.879; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Foreign uncertainty index: Level Chi2 Stat.: 2,175.163; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- GDP per capita (log): Level Chi2 Stat.: 277.864; Level Prob. > Chi2: 0.999; First difference Chi2 Stat.: 11,639.629; First difference Prob. > Chi2: 0.000; Conclusion: I(1)
- Foreign GDP per capita (log): Level Chi2 Stat.: 231.969; Level Prob. > Chi2: 1.000; First difference Chi2 Stat.: 11,098.391; First difference Prob. > Chi2: 0.000; Conclusion: I(1)
- Exchange rate volatility: Level Chi2 Stat.: 5,587.854; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Total migrant stock (log): Level Chi2 Stat.: 166.863; Level Prob. > Chi2: 1.000; First difference Chi2 Stat.: 10,260.483; First difference Prob. > Chi2: 0.000; Conclusion: I(1)
- Financial institutions access index: Level Chi2 Stat.: 225.1843; Level Prob. > Chi2: 1.000; First difference Chi2 Stat.: 9,427.435; First difference Prob. > Chi2: 0.000; Conclusion: I(1)
- Inflation: Level Chi2 Stat.: 1846.77; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Monetary policy rate: Level Chi2 Stat.: 1,176.934; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Real GDP growth: Level Chi2 Stat.: 2,184.565; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Foreign GDP growth: Level Chi2 Stat.: 3,070.572; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Remittances costs (500 $): Level Chi2 Stat.: 542.497; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Private investment ratio to GDP: Level Chi2 Stat.: 524.845; Level Prob. > Chi2: 0.000; Conclusion: I(0)
- Public spending on heath per capita: Level Chi2 Stat.: 224.890; Level Prob. > Chi2: 1.000; First difference Chi2 Stat.: 7,225.490; First difference Prob. > Chi2: 0.000; Conclusion: I(1)
- Public spending on education per capita: Level Chi2 Stat.: 115.250; Level Prob. > Chi2: 1.000; First difference Chi2 Stat.: 4,962.610; First difference Prob. > Chi2: 0.000; Conclusion: I(1)

### Key takeaways and policy implications
- Core findings:
  - Foreign (host-country) uncertainty has a robust negative effect on remittance flows; migrants reduce remittances contemporaneously when host-country uncertainty rises.
  - Domestic uncertainty has an ambiguous average effect; net impact depends on balance between altruistic consumption-smoothing motives and investment motives.
  - High private investment-to-GDP countries tend to experience remittance declines when domestic uncertainty rises (investment channel dominates).
  - Low public spending on education and health is associated with remittances rising with domestic uncertainty (altruism/insurance channel dominates).
- Policy-relevant implications:
  - Maintain stability in remittance-sending countries: reduce labor-market uncertainty and expand social safety nets for migrants to stabilize remittance flows.
  - Receiving-country interventions should be targeted: increasing public provision of health and education can reduce household reliance on remittances for basic services and reorient remittances toward investment.
  - Foster a conducive investment climate (political stability, regulatory clarity, ease of doing business) to attract remittance-funded productive investment.
  - Reduce remittance transaction costs to support remittance resilience (transaction costs inversely related to remittance volumes).

*Source: wpiea2024244-print-pdf - Introduction (IMF working paper PDF contents list).*

### Introduction ...........................................................................................................

### wpiea2024244-print-pdf - Introduction ...........................................................................................................

### Introduction and Scope
- The document contains an Introduction section followed by focused analysis on the links between uncertainty and remittances.
- Main thematic areas covered:
  - "Uncertainty and Remittances: How Close are the Links?"
  - "Empirical Strategy"
  - "Remittance and Uncertainty Data and their Measurements"
  - "The Model, Sample and Econometric Approach"
  - "Empirical Findings"
  - "Conclusion"

### Empirical Strategy and Data Components
- Empirical strategy is organized across distinct methodological subsections:
  - Empirical Strategy
  - Remittance and Uncertainty Data and their Measurements
  - The Model, Sample and Econometric Approach
- Annexed materials and appendices supporting empirical work include:
  - Country Sample Composition
  - Summary statistics and correlation matrix
  - Correlation matrix
  - Variable Definitions and Sources
  - Stationarity tests: Maddala and Wu (1999)

### Empirical Findings: Structure of Results
- Empirical Findings are decomposed into:
  - Fixed-Effect Estimations
  - The Local Projection Estimations
- Additional robustness and mechanism explorations are indicated by listed tables and figures (titles retained exactly as listed in the source).

### Figures (as listed in the source)
- 1. Trends in the Economic Uncertainty Index, 1990Q1 to 2019Q4
- 2. Economic Uncertainty Index across Developing Countries, 2010Q1 to 2019Q4
- 3. Trends in Quarterly Remittance Flows, 1999Q1–1999Q4
- 4. Correlation Between Economic Uncertainty and Remittances
- 5. Marginal Impact of Domestic Uncertainty on Remittances Conditional to the Country’s Private Investment Ratio to GDP and Government Spending on Health and Education per Capita.
- 6. IRF of Remittances with Respect to Domestic Uncertainty
- 7. IRF of Remittances with Respect to Foreign Uncertainty
- 8. IRF of Remittances with Respect to Domestic Uncertainty and Conditional to the Private Investment Ratio to GDP
- 9. IRF of Remittances with Respect to Domestic Uncertainty and Conditional to Public Health Spending per Capita
- 10. IRF of Remittances with Respect to Domestic Uncertainty and Conditional to Public Education Spending per Capita

### Tables (as listed in the source)
- 1. Uncertainty and Remittances
- 2. Uncertainty and Remittances: Alternative Indicators of Remittances
- 3. Uncertainty and Remittances: Additional Control Variables
- 4. Dealing with Endogeneity using an Instrumental Variable Approach
- 5. Domestic Uncertainty and Remittances: The Role of Private Investment and Public Human Capital Spending

### Appendix and Annex Materials (as listed)
- APPENDIX TABLE 1. First-Stage Regressions of the Instrumental Variable Approach
- ANNEXES:
  - 1. Country Sample Composition
  - 2. Summary statistics and correlation matrix
  - 3. Correlation matrix
  - 4. Variable Definitions and Sources
  - 5. Stationarity tests: Maddala and Wu (1999)

*Source: wpiea2024244-print-pdf - Introduction (IMF working paper PDF contents list).*

### Introduction

### Introduction

### Background and motivation
- Remittance flows are influenced by altruistic motives and self-interest considerations (Amuedo-Dorantes, 2014; Ratha, 2017).
- Existing empirical literature largely omits explicit measurement of uncertainty, focusing instead on measurable risk factors such as economic growth volatility and exchange rate volatility (Backer et al., 2013; Fernández-Villaverde et al., 2011; Leahy et al., 1996).
- Advances in quantifying uncertainty (e.g., Bloom, 2009) and text-mining of "big data" now permit construction of uncertainty indices, allowing investigation of how uncertainty affects remittances.

### Why uncertainty matters for remittances
- Channels through which uncertainty can affect remittances:
  - Host-country (foreign) uncertainty: economic downturns, unemployment, exchange rate fluctuations, immigration policy changes → reduce migrants’ income and increase precautionary savings, lowering remittances.
  - Home-country (domestic) uncertainty: political turmoil, conflicts, disasters → may increase demand for remittances for consumption smoothing (altruistic channel) or reduce remittances intended for investment if expected returns fall (investment channel).
- Behavioral mechanisms include the 'option value of waiting' under ambiguity, and migrants’ limited access to social safety nets increasing precautionary motives.

### Research aim and contribution
- Objective: examine the impact of uncertainty on remittance flows, considering both origin (domestic) and destination (foreign) countries.
- Novel features:
  - Uses a new quarterly remittance dataset matched to the Ahir, Bloom, and Furceri (2022) uncertainty indices.
  - Sample: 77 developing countries during 1999Q1─2019Q4.
  - Employs fixed-effect estimations and local projection (Jordà, 2005) dynamic analysis.
  - Disentangles altruism (countercyclical remittances) and investment channels (procyclical remittances) to explain heterogeneous domestic responses.

### Data and measurements
- Uncertainty index:
  - Source: Ahir, Bloom and Furceri (2022), constructed from EIU quarterly country reports (mentions of “uncertainty” normalized by total word count).
  - Coverage: 143 countries, 1952Q1–2021Q2 (used series for developing countries 1995Q1–2020Q1 in figures).
- Remittances:
  - Novel quarterly dataset for 95 countries (18 high-income, 62 middle-income, 15 low-income) building on Kpodar et al. (2023).
  - Dataset spans from 1971Q1 (a few countries) through 2020Q4 for most countries; consolidated at the receiving country level (not bilateral).
  - Definition follows standard remittance definition: personal transfers and compensation of employees; when missing, workers' remittances or omission of employer compensation used as proxies.
- Final empirical sample for main analysis: 77 developing countries, 1999Q1–2019Q4.
- Key descriptive statistics (developing-country sample, quarterly):
  - Median remittances per country ≈ USD 0.5bn per quarter (1 percent of GDP).
  - 75th percentile: > USD 1.8bn per quarter (2.7 percent of GDP).
  - 25th percentile: ≤ USD 0.1bn per quarter (0.4 percent of GDP).
- Control variables include: domestic and foreign GDP per capita, total migrant stock, domestic currency per U.S. dollar (level and volatility), exchange rate volatility (quarterly standard deviation).

### Empirical strategy and baseline model
- Baseline linear model (Eq(1)): log(Rem)_{c,t} = γ + α Uncert_{c,t}^l + β Uncert_{c,t}^f + Θ X_{c,t} + μ_c + ϑ_t + ε_{c,t}
  - Rem_{c,t}: remittances in millions of US dollars (country c, time t).
  - Uncert^l: domestic (local) economic uncertainty.
  - Uncert^f: foreign (host-country) economic uncertainty (proxied by migrant-weighted average across host countries).
  - X: controls (income per capita of receiving and sending countries, migrant stock, exchange rate level and volatility).
  - Country fixed effects and time dummies (year and quarter) included; standard errors robust to heteroscedasticity.
- Dynamic analysis: Local projection (LP) specification (Eq(2)) with n = 4 lags (quarterly data) and forecast horizon h up to 5 quarters; includes Teulings and Zubanov (2014) correction to control for subsequent shocks.

### Main empirical findings — fixed-effect estimations
- Domestic uncertainty:
  - Baseline: domestic uncertainty does not significantly influence remittance magnitudes across multiple specifications (columns 1, 3, 4, and 6 in Table 1).
  - Heterogeneity: net effect of domestic uncertainty can be positive or negative depending on country characteristics—domestic uncertainty tends to increase remittances where investment motives are weak and public provision of health and education is poor; it reduces remittances where remittances are channeled to investment.
- Foreign (host-country) uncertainty:
  - Robust negative effect on remittances across specifications.
  - Representative coefficients from Table 1:
    - Foreign uncertainty index coefficients: -0.238 [0.102]**, -0.223 [0.084]***, -0.225 [0.076]***, -0.225 [0.086]** (columns 2, 3, 5, 6).
- Control variables (selected, Table 1):
  - GDP per capita (domestic) positive and significant: coefficient values include 1.271 [0.321]***, 1.090 [0.301]***, 1.260 [0.320]***, 1.250 [0.305]*** (columns 1–6).
  - Foreign GDP per capita (log) positive and significant where included: 0.390 [0.093]***, 0.449 [0.083]***, 0.385 [0.093]***.
  - Domestic currency per U.S. dollar (log) positive: 0.237 [0.109]**, 0.184 [0.111], 0.242 [0.110]**.
  - Exchange rate volatility negative: -0.384 [0.209]*, -0.418 [0.240]*, -0.387 [0.207]*.
- Robustness:
  - Alternative remittance measures (Table 2): results consistent — domestic uncertainty non-significant; foreign uncertainty negative (e.g., -0.424 [0.217]* for remittances per migrant).
  - Additional controls (Table 3): inclusion of financial access, inflation, monetary policy rate, GDP growth, remittance transaction costs does not alter main results; foreign uncertainty remains negative and significant (examples: -0.209 [0.092]**, -0.224 [0.086]**, -0.172 [0.070]**).
  - Instrumental variables (Table 4): domestic uncertainty instrumented by presidential/legislative election dummies and export-weighted partner uncertainty — exogenous component of domestic uncertainty remains statistically insignificant for remittances while foreign uncertainty stays negative (e.g., -0.229 [0.092]**).

### Heterogeneity: disentangling altruism vs. investment channels
- Investment channel:
  - Hypothesis: higher private investment-to-GDP ratio → larger share of remittances for investment → domestic uncertainty reduces remittances via lower expected returns on investment.
  - Evidence (Table 5): lagged private investment-to-GDP ratio positive and significant (e.g., 0.012 [0.004]***, 0.016 [0.004]***); interaction Inv*Domestic uncertainty negative and significant (-0.025 [0.010]**, -0.021 [0.009]**), supporting the investment-deterrence channel.
- Altruism / public service channel:
  - Hypothesis: in countries with low public spending on health and education, domestic uncertainty elicits higher remittances to buffer households.
  - Evidence (Table 5): interaction terms of per capita public health and education spending with domestic uncertainty are negative and significant (e.g., Health*Domestic uncertainty [0.000]**, Educ*Domestic uncertainty [0.003]** in relevant columns), implying remittances respond more positively to domestic uncertainty where public provision is weak.
- Conditional results and marginal effects:
  - Using column 9 specification to compute country-level marginal impacts (Figure 5): for most sample countries, net impact of domestic uncertainty on remittances is positive; for a subset (high private investment contexts), remittances decline with rising domestic uncertainty.
  - LP-based threshold analysis (Figure 8): marginal impact of domestic uncertainty on remittances crosses zero at a private investment ratio of 18 percent of GDP; sample median = 15.14 percent; sample average = 16 percent.

### Dynamics — Local projection (LP) results
- LP impulse response functions (IRFs):
  - Domestic uncertainty: LP shows no discernible contemporaneous correlation with remittances, consistent with fixed-effect estimations once dynamics are considered (Figure 6).
  - Foreign uncertainty: LP shows a significant negative effect on remittances in the initial quarter following a foreign uncertainty shock; marginal effect attenuates and becomes statistically insignificant thereafter (Figure 7).
  - Comparison: fixed-effect estimates display a stronger immediate negative response to foreign uncertainty than LP estimates, suggesting potential overestimation when dynamic effects are ignored.
- LP heterogeneity:
  - Conditional IRFs at the 10th and 90th percentiles of the private investment distribution:
    - At low private investment (10th percentile), domestic uncertainty is associated with higher remittances in the first quarter only, with effects dissipating after quarter 1.
    - At high private investment (90th percentile), domestic uncertainty yields a negative and significant effect in quarter 1.
    - The marginal impact crosses zero at private investment ratio = 18 percent of GDP.

### Robustness to endogeneity and additional checks
- Instruments used:
  - Domestic election dummies (presidential and legislative) and export-weighted partner uncertainty; first-stage shows elections raise domestic uncertainty.
  - IV fixed-effect and IV-LP analyses leave core conclusions intact: domestic uncertainty not robustly linked to remittances; foreign uncertainty reduces remittances.
- Additional checks:
  - Results robust to use of growth volatility measures (standard deviation of annual real GDP growth over preceding 5 years), inclusion of remittance costs, and alternative remittance indicators (levels, per capita, ratios).

### Key takeaways and policy-relevant implications
- Main empirical findings:
  - Foreign (host-country) uncertainty has a robust negative effect on remittance flows; migrants reduce remittances contemporaneously when host-country uncertainty rises.
  - Domestic (home-country) uncertainty has an ambiguous average effect; the net impact depends on the relative strength of altruistic consumption-smoothing motives versus investment motives.
  - Countries with high private investment-to-GDP ratios tend to experience declines in remittances when domestic uncertainty rises (investment channel dominates).
  - Countries with low public spending on health and education tend to see remittances increase with domestic uncertainty (altruism/insurance channel dominates).
- Practical implications:
  - Policymakers in remittance-receiving countries should consider the composition of remittance usage (consumption vs. investment) when assessing vulnerability to external shocks and uncertainty.
  - Improving public provisioning of education and health can alter remittance dynamics by reducing households’ reliance on remittances for basic services.
  - Host-country policies that reduce labor-market uncertainty or expand social safety nets for migrants could stabilize remittance flows to origin countries.
  - Transaction costs matter: reducing remittance costs can support remittance resilience, as remittance costs are inversely related to remittance volumes.

*Source: wpiea2024244-print-pdf - Introduction*

### 6.1 percent

### 6.1 percent

### Remittances response conditional on private investment
- In countries with high private investment ratios, remittances decline with rising domestic uncertainty (illustrated by the 90th percentile: Private Investment Ratio to GDP = 25.3 percent).
- The IRFs for private investment show a quick and contained reaction within the first quarter.
- Shaded area in the referenced figures represents the 90 percent confidence interval.
- Source: Authors’ calculations.

### Remittances response conditional on public health spending per capita
- A low level of public spending on health per capita is associated with a positive response of remittances to a rise in uncertainty.
- This relationship reverses for countries with higher public health spending in relation to their population.
- Possible interpretation: out-of-pocket expenses incurred by households are lower where public health spending is high, so migrants’ families may not need to cut back health spending amid adverse economic shocks.
- The IRFs conditional on public health spending are slower and more prolonged compared to private investment:
  - For public health spending, the marginal impact becomes significant only beyond the first quarter.
  - In countries with high public health spending, remittances decline only after the fourth quarter.
  - In countries with low public health spending, the positive response occurs from the second quarter after the uncertainty shock.
- Example numeric percentiles from Figure 9:
  - 10th percentile (Public Health Spending per Capita = US$22.5)
  - 90th percentile (Public Health Spending per Capita = US$524.0)
- Notes: Shaded area represents the 90 percent confidence interval. Source: Authors’ calculations.

### Remittances response conditional on public education spending per capita
- IRFs conditional on public education spending per capita exhibit similar patterns as for public health spending per capita:
  - Remittances increase with uncertainty in countries with inadequate public investment in education.
  - Remittances decline with uncertainty (albeit with a lag) where public investment in education is high.
- Example numeric percentiles from Figure 10:
  - 10th percentile (Public Education Spending per Capita = US$1.1)
  - 90th percentile (Public Education Spending per Capita = US$35.4)
- Notes: Shaded area represents the 90 percent confidence interval. Source: Authors’ calculations.

### Main conclusions from the paper
- Uncertainty is a key driver of remittances, with significant implications for policymaking.
- Sample and estimation:
  - Sample: 77 countries from 1999Q1 2019Q4.
  - Econometric estimations reveal that uncertainty in the remittance-sending country discourages remittances (migrants build up precautionary savings).
- Receiving-country effects are heterogeneous and non-linear:
  - In countries with high private investment ratios, remittances decline with rising domestic uncertainty.
  - In countries with low public spending on education and health, remittances react positively to domestic uncertainty, acting as a social safety net mechanism.
- Robustness:
  - Results remain robust to alternative indicators of remittances, inclusion of additional control variables, treatment of potential endogeneity concerns, and different econometric methodologies (fixed effect estimates and local projections).

### Policy implications
- Maintain stability in remittance-sending countries:
  - The negative relationship between remittance flows and uncertainty in sending countries highlights the importance of maintaining a stable economic and political environment.
  - Policies that provide migrants with stable employment opportunities and access to social safety nets can help mitigate adverse effects of uncertainty on remittances.
  - Policies aimed at reducing volatility and enhancing predictability in financial markets could bolster migrants’ confidence to sustain or increase remittance flows.
- Targeted receiving-country interventions:
  - The ambiguous effect of receiving-country uncertainty on remittances—driven by investment disincentives and altruistic familial support—requires nuanced, targeted policy interventions.
  - Increasing public investment in healthcare and education could enable households to direct a greater share of remittances towards investment.
  - Fostering a conducive investment climate in remittance-receiving countries (political stability, regulatory clarity, ease of doing business) could attract remittance flows into productive investments, catalyzing economic growth and job creation.

### Key statistics and sample summaries (selected figures from Annexes)
- Sample: 77 countries (1999Q1 2019Q4).
- Annex 2 summary statistics (selected rows, exact values preserved):
  - Remittances (millions of US dollars, log): Obs 4,188; Mean 5.408; Std. dev. 2.042; Min -1.738; Max 10.030
  - Remittances per migrant (thousands of US dollars, log): Obs 4,188; Mean -1.256; Std. dev. 1.548; Min -7.412; Max 8.226
  - Remittances to GDP (log): Obs 4,188; Mean -0.299; Std. dev. 1.634; Min -6.695; Max 2.428
  - Domestic uncertainty index: Obs 3,580; Mean 0.200; Std. dev. 0.215; Min 0.000; Max 2.038
  - Foreign uncertainty index: Obs 4,188; Mean 0.219; Std. dev. 0.147; Min 0.001; Max 1.668
  - Private investment ratio to GDP: Obs 3,538; Mean 17.838; Std. dev. 6.880; Min -7.019; Max 54.799
  - Public spending on heath per capita (US dollars): Obs 3,926; Mean 223.272; Std. dev. 202.781; Min 8.362; Max 1192.82
  - Public spending on education per capita (US dollars): Obs 2,942; Mean 16.084; Std. dev. 15.230; Min 0.479; Max 86.432
  - Public spending on health and education per capita (US dollars): Obs 2,856; Mean 241.890; Std. dev. 223.231; Min 9.264; Max 1242.15

*Source: Authors’ calculations as presented in the chapter.*

### Annex 4. Variable Definitions and Sources

### Annex 4. Variable Definitions and Sources

### Variable definitions and data sources
- Remittances
  - Definition: Sum of personal transfers and compensation of employees in millions USD/Local Currency.
  - Source: Central Banks and National institutes of Statistics (Kpodar et al., 2023)

- Domestic uncertainty index
  - Definition: Index calculated as the number of occurrences of the word uncertainty and its equivalents over the total number of words in the Economist Intelligence Unit (EIU) report
  - Source: Ahir and Furceri (2022)

- Foreign uncertainty index
  - Definition: Weighted average of the uncertainty index of all host countries with the weight being the host country’s share of total migrants
  - Source: Authors' calculations using Ahir and Fuceri (2022) database, World Development Indicators (WDI) and World Bank bilateral migrant stock data.

- Foreign GDP per capita (log)
  - Definition: Weighted average of GDP per capita of all host countries (in log) with the weight being the host country’s share of total migrants

- Foreign GDP growth
  - Definition: Weighted average of annual GDP growth rate of all host countries with the weight being the host country’s share of total migrants

- Exchange rate volatility
  - Definition: Standard deviation of domestic currency per U.S. Dollar (log)
  - Source: Authors' calculations using the International Financial Statistics (IFS)

- Total migrant stock (log)
  - Definition: Total migrant stock of the receiving country in millions (in log)
  - Source: World Bank

- Remittances costs (500 $)
  - Definition: Transaction cost (in percent of a $500 remittance, log) for the home country

- Financial institutions access index
  - Definition: Financial Access Index calculated using the number of bank branches and ATMs per 100,000 adults. The index ranges from 0 to 1.
  - Source: International Monetary Fund (IMF)

- Monetary policy rate
  - Definition: Quarterly average of the main monetary policy rate of the Central Bank of the receiving country
  - Source: International Financial Statistics (IFS)

- Private investment ratio to GDP
  - Definition: Private gross fixed capital formation, current prices ( % GDP)
  - Source: World Economic Outlook (WEO) database

- GDP per capita (log)
  - Definition: GDP per capita (constant 2010 US$) in log
  - Source: World Development Indicators (WDI)

- Real GDP growth
  - Definition: Annual GDP growth rate

- Inflation
  - Definition: Annual percentage change in the consumer price index.

- Public spending on heath per capita
  - Definition: Current government health expenditure per capita (current US$) in the receiving country

- Public spending on education per capita
  - Definition: Government expenditure on education per capita in receiving country

- Legislative elections
  - Definition: Dummy variable equal to 1 for the period of legislative elections and zero otherwise.
  - Source: Database of Political Institutions (DPI2020)

- Presidential elections
  - Definition: Dummy variable equal to 1 for the period of presidential elections and zero otherwise.

- Bilateral export
  - Definition: Bilateral export of goods (in thousands current US$)
  - Source: Direction of Trade and Statistics (DOTS), IMF

*Source: authors*

### Annex 5. Stationarity tests: Maddala and Wu (1999) — Results and conclusions
- Table columns: Variable | Level Chi2 Stat. | Level Prob. > Chi2 | First difference Chi2 Stat. | First difference Prob. > Chi2 | Conclusion

- Remittances (log)
  - Level Chi2 Stat.: 348.627
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Remittances per migrant (log)
  - Level Chi2 Stat.: 352.677
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Remittances to GDP (log)
  - Level Chi2 Stat.: 550.842
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Remittances per capita (log)
  - Level Chi2 Stat.: 420.283
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Remittances per migrant as a share of income per capita of the host economy (log)
  - Level Chi2 Stat.: 444.437
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Domestic uncertainty index
  - Level Chi2 Stat.: 2,951.879
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Foreign uncertainty index
  - Level Chi2 Stat.: 2,175.163
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- GDP per capita (log)
  - Level Chi2 Stat.: 277.864
  - Level Prob. > Chi2: 0.999
  - First difference Chi2 Stat.: 11,639.629
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

- Foreign GDP per capita (log)
  - Level Chi2 Stat.: 231.969
  - Level Prob. > Chi2: 1.000
  - First difference Chi2 Stat.: 11,098.391
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

- Exchange rate volatility
  - Level Chi2 Stat.: 5,587.854
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Total migrant stock (log)
  - Level Chi2 Stat.: 166.863
  - Level Prob. > Chi2: 1.000
  - First difference Chi2 Stat.: 10,260.483
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

- Financial institutions access index
  - Level Chi2 Stat.: 225.1843
  - Level Prob. > Chi2: 1.000
  - First difference Chi2 Stat.: 9,427.435
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

- Inflation
  - Level Chi2 Stat.: 1846.77
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Monetary policy rate
  - Level Chi2 Stat.: 1,176.934
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Real GDP growth
  - Level Chi2 Stat.: 2,184.565
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Foreign GDP growth
  - Level Chi2 Stat.: 3,070.572
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Remittances costs (500 $)
  - Level Chi2 Stat.: 542.497
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Private investment ratio to GDP
  - Level Chi2 Stat.: 524.845
  - Level Prob. > Chi2: 0.000
  - First difference: (no first-difference statistics reported)
  - Conclusion: I(0)

- Public spending on heath per capita
  - Level Chi2 Stat.: 224.890
  - Level Prob. > Chi2: 1.000
  - First difference Chi2 Stat.: 7,225.490
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

- Public spending on education per capita
  - Level Chi2 Stat.: 115.250
  - Level Prob. > Chi2: 1.000
  - First difference Chi2 Stat.: 4,962.610
  - First difference Prob. > Chi2: 0.000
  - Conclusion: I(1)

*Source: authors*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024244-print-pdf.pdf_