## _wp14144

## Source details

**Canonical URL:** [_wp14144](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14144.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14144.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14144.pdf.json)

---

### I. Introduction
- Remittances to developing countries have grown significantly; literature on macroeconomic and social impacts exists but effects on credit quality (NPLs) have been little studied.
- Paper analyzes impact of remittances on nonperforming loans (NPLs) using panel data of 141 developing countries over 2000–2011 and various estimation techniques.
- Main empirical finding: remittances are negatively correlated with NPLs in the sample, suggesting the income-stabilizing effect dominates the risk-inducing effect.

### II. Theoretical background: channels linking remittances and NPLs
- Two opposing theoretical channels:
  - Risk-inducing effect: remittances increase banks’ loanable funds and credit to private sector, potentially increasing risk and worsening credit quality.
  - Income-stabilizing effect: remittances enhance borrowers’ capacity to repay (stable transfers can act as collateral), reduce income volatility, and improve banks’ ability to screen borrowers.
- Institutional and bank-specific factors affect credit quality: diversification, capitalization, bank size, public ownership, supervisory framework, and efficiency.
- Non-linearity: growth impact of remittances is maximized with good institutions; self-reinforcing mechanism implies stronger remittance effects in countries with higher institutional quality and financial development.
- Caution: increased income could also spur excess risk-taking by banks; credit risk may be limited if new loans are extended mainly to remittance receivers.

### III. Empirical framework and data
- Model specifications:
  - Linear: NPLi,t = θRi,t + φσi,t + X'i,tβ + ui + εi,t
    - NPLi,t: NPL ratio; Ri,t: remittances-to-GDP; σi,t: output growth volatility (five-year rolling standard deviation of real GDP growth per capita); X: traditional determinants including real GDP growth, GDP per capita, exports-to-GDP, credit boom dummy; governance index (PCA composite scaled 0–100); ui: country fixed effects.
  - Non-linear / stabilizing channel: NPLi,t = θ1 + θ2 σi,t × Ri,t + φσi,t + X'i,tβ + ui + εi,t (interaction between remittances and output volatility).
- Estimation methods:
  - OLS with country fixed effects.
  - Fractional logit (quasi-likelihood) to account for bounded dependent variable between zero and one and values of zero or one.
- Sample and data:
  - Unbalanced panel of 141 low- and middle-income countries over 2000–2011.
  - Remittances definition: World Bank aggregate (includes workers’ remittances and compensation of employees when not separable).
  - Credit boom defined following Mendoza and Terrones (2008): cyclical component of credit > 1.75 × its standard deviation (HP filter used).

### IV. Results — Baseline estimates and robustness
- Traditional determinants (selected coefficients; OLS-FE and Fractional Logit reported where noted):
  - Output growth volatility: OLS-FE 0.439; Fractional Logit 0.335.
  - Credit boom: OLS-FE 0.010; Fractional Logit 0.013.
  - Ln GDP per capita: OLS-FE -0.187; Fractional Logit -0.194.
  - Real GDP growth: OLS-FE -0.317; Fractional Logit -0.327.
  - Exports-to-GDP and Governance index: not significant in baseline FE regressions.
  - Sample stats: Observations 638; Number of countries 71; R2 0.32 (OLS-FE) and 0.67 (Fractional Logit).
- Linear remittances effects (coefficients across specifications and methods):
  - Reported remittances-to-GDP coefficients: -0.517, -0.438, -0.515, -0.364, -0.25, -0.354 (all negative and statistically significant across specifications).
  - Quantification: an increase in remittance-to-GDP by one half standard-deviation in the sample (4 percent of GDP) is associated with a drop in the NPL ratio of about 1–2 percentage points, ceteris paribus.
  - Controls maintain expected signs and significance.
  - Sample stats for Table 2: Observations 624; Number of countries 70; R2 ranges 0.32–0.35 (OLS-FE) and 0.67–0.68 (Fractional Logit).
- Non-linearity / stabilizing channel (interaction results):
  - Additive remittances-to-GDP: OLS-FE -0.413; Fractional Logit -0.239 (both significant).
  - Output growth volatility: OLS-FE 0.714; Fractional Logit 0.634 (both significant).
  - Interaction Remittances-to-GDP × Output growth volatility: OLS-FE -1.911; Fractional Logit -1.715 (both significant) — remittances dampen the adverse effect of volatility on NPLs.
  - Joint significance of non-linearity: P-value 0.0002 (OLS-FE) and 0.0000 (Fractional Logit).
  - Reported remittances threshold (in percent of GDP): 35.5 (Table 3, OLS-FE reported value).
  - Sample stats for Table 3: Observations 624; Number of countries 70; R2 0.36 (OLS-FE) and 0.68 (Fractional Logit).

### V. Endogeneity concerns and instrumental variables strategy
- Endogeneity sources:
  - Reverse causality: countercyclicality of remittances when migrants insure those left behind, potentially biasing estimates.
  - Redundancy effect: balance of payments remittance flows may reflect countries with certain financial development levels.
  - Omitted variable bias: unobserved time-varying factors correlated with both NPLs and remittance inflows.
- Instrumental variables:
  - Primary instrument: real per capita GDP growth in remittance-sending countries, computed as the weighted average of GDP growth rates in potential migrant destination countries with weights equal to bilateral migration shares (World Bank Bilateral Migration Database).
  - Complementary instrument: two-year lag of the remittance variable (remittances-to-GDP in t-2).
  - Identification assumption: better economic conditions in migrant-destination countries are associated with larger remittances to sender countries, conditional on fixed effects and recipient-country conditions.
  - Diagnostics: first-stage strength tests and Hansen over-identification tests.

### VI. Instrumental Variable results — Linear and non-linear specifications
- Linear IV results (Table 4; second-stage coefficients (robust standard error); significance):
  - Sample: Period 2000–11; Observations 560; Number of countries 65; Country fixed-effects: Yes.
  - Remittances-to-GDP: -0.636 (0.183)*** 
  - Output growth volatility: 0.380 (0.130)*** 
  - Credit boom: 0.006 (0.004) 
  - Ln GDP per capita: -0.199 (0.016)*** 
  - Real GDP growth: -0.295 (0.061)*** 
  - Exports-to-GDP: -0.002 (0.049) 
  - Governance index: -0.003 (0.011)
- Linear IV first-stage indicators:
  - Growth in migrant host countries: 0.09 (0.041)** 
  - Remittances-to-GDP (in t-2): 0.493 (0.060)*** 
  - F-statistic of remittances-to-GDP Instrumental equation: 54.2
  - Hansen over-identification test (p-value): 0.57
- Interpretation:
  - IV estimates confirm a negative and robust association between remittances and declines in NPL ratios.
  - IV point estimates in absolute terms are higher than OLS counterparts, suggesting positive endogeneity bias in OLS.
- Non-linear IV results (Table 5; second-stage coefficients (robust standard error); significance):
  - Sample: Period 2000–11; Observations 560; Number of countries 65; Country fixed-effects: Yes.
  - Remittances-to-GDP: -0.485 (0.172)*** 
  - Remittances-to-GDP * output growth volatility: -1.301 (0.728)* 
  - Output growth volatility: 0.539 (0.197)*** 
  - Credit boom: 0.005 (0.004) 
  - Ln GDP per capita: -0.201 (0.015)*** 
  - Real GDP growth: -0.285 (0.063)*** 
  - Exports-to-GDP: -0.008 (0.048) 
  - Governance index: -0.003 (0.011)
- Non-linear IV first-stage and diagnostics (selected):
  - Growth in migrant host countries: 0.121 (0.036)*** ; 0.002 (0.001)
  - Remittances-to-GDP (in t-2): 0.514 (0.069)*** ; -0.005 (0.003)* 
  - Remittances-to-GDP (in t-2)*output growth volatility: -0.196 (0.157) ; 0.568 (0.008)*** 
  - F-statistic of instrumentation equation: 31.6 ; 2624.0
  - Hansen over-identification test (p-value): 0.66
  - Joint significance of the non-linearity: P-val 0.0001
- Interpretation:
  - Instrumentation does not alter previous findings; remittances reduce NPLs and this stabilizing role is stronger in countries with high macroeconomic volatility.

### VII. Robustness checks and alternative measures
- Alternative remittance measure:
  - Remittances-to-deposits used as alternative; Appendix results confirm robust and negative impact of remittances on NPLs.
  - Table A3 Remittances-to-deposit coefficients (standard errors):
    - (1) -0.020 (0.011)* 
    - (2) -0.020 (0.007)*** 
    - (3) -0.019 (0.007)*** 
    - (4) -0.019 (0.007)*** 
    - (5) -0.017 (0.008)** 
  - Table A3 additional stats:
    - # Countries: 59
    - R2: 0.03, 0.37, 0.38, 0.38, 0.40 across columns (1)–(5)
    - N: 475, 475, 473, 473, 473
- Sample bias sensitivity:
  - Splitting the sample by income groups (low income, lower middle, and upper middle income) broadly confirms previous results (detailed results available upon request).
- Descriptive statistics (Table A1):
  - NPL ratio: Obs 624; Mean 0.09; Std. Dev. 0.08; Min 0.00; Max 0.57
  - Remittances-to-GDP: Obs 624; Mean 0.05; Std. Dev. 0.07; Min 0.00; Max 0.51
  - Output growth volatility: Obs 624; Mean 0.03; Std. Dev. 0.03; Min 0.00; Max 0.34
  - Credit boom: Obs 624; Mean 0.43; Std. Dev. 0.50; Min 0.00; Max 1.00
  - ln GDP per capita: Obs 624; Mean 7.48; Std. Dev. 0.96; Min 5.03; Max 9.36
  - Real GDP growth: Obs 624; Mean 0.05; Std. Dev. 0.04; Min -0.18; Max 0.27
  - Exports-to-GDP: Obs 624; Mean 0.37; Std. Dev. 0.18; Min 0.07; Max 1.20
  - Governance index: Obs 624; Mean 5.91; Std. Dev. 1.35; Min 2.79; Max 10.00
- Descriptive and regional statistics:
  - Time frame: 2000–2011.
  - Sample: 141 developing countries; key estimation samples report 70–71 countries and 624–638 observations.
  - Remittances trend: average remittances to developing countries rose from $700 million (1990s average) to $2.8 billion in 2010.
  - Regional average NPLs (2000–11 sample averages):
    - MENA: 13.9 percent
    - SAS: 12.8 percent
    - SSA: 11.0 percent
    - EAP: 10.9 percent
    - ECA: 9.0 percent
    - LAC: 5.2 percent
  - Top regional remittance recipients (2007–11 average, percent of GDP), selected:
    - SSA: Lesotho (35.5), Togo (10.5), Senegal (10.5), Cape Verde (9), Gambia (8.7)
    - EAP: Tonga (23.6), Samoa (20.2), Philippines (10.9), Vietnam (7.4), Fiji (4.6)
    - SAS: Nepal (21), Bangladesh (10.9), Sri Lanka (8), Pakistan (5), India (3.5)
    - MENA: Lebanon (21.2), Jordan (15.7), Morocco (7.5), Egypt (5.4), Yemen (5.1)
    - LAC: Haiti (21.3), Honduras (18.8), El Salvador (16.9), Jamaica (15.4), Guyana (15.1)

### VIII. Key empirical findings and policy implications
- Core empirical findings:
  - Remittances are robustly negatively associated with NPL ratios in the sample of developing countries (point estimates for remittances-to-GDP range from -0.25 to -0.517 in OLS specifications and as low as -0.636 in IV).
  - Higher output growth volatility is positively associated with larger NPL ratios (reported coefficients include 0.335, 0.439, 0.481, 0.714, 0.634).
  - Credit booms are associated with higher NPLs (coefficients around 0.006–0.013 across specifications).
  - Interaction results: remittances reduce the marginal impact of macroeconomic volatility on NPLs (interaction coefficients -1.911, -1.715, -1.301).
  - Non-linearities are econometrically significant (joint P-values 0.0002, 0.0000, 0.0001).
- Interpretation and implications:
  - Income-stabilizing effects of remittances appear to dominate the risk-inducing effects in the sample, implying remittances enhance borrowers’ capacity to repay and act as shock absorbers.
  - Countries with high macroeconomic volatility benefit more from remittances in terms of banking stability.
  - Policy relevance: remittances can reduce NPLs and support the development of a sound formal financial sector, which is critical for sustaining high economic growth.

*IMF Working Paper content (excerpt provided).*

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

### _wp14144 - References

### I. Introduction
- Remittances to developing countries have grown significantly; literature on macroeconomic and social impacts exists but effects on credit quality (NPLs) have been little studied.
- Paper analyzes impact of remittances on nonperforming loans (NPLs) using panel data of 141 developing countries over 2000–2011 and various estimation techniques.
- Main empirical finding reported: remittances are negatively correlated with NPLs in the sample, suggesting the income-stabilizing effect dominates the risk-inducing effect.

### II. Theoretical background: channels linking remittances and NPLs
- Two opposing theoretical channels:
  - Risk-inducing effect: remittances increase banks’ loanable funds and credit to private sector, potentially increasing risk and worsening credit quality.
  - Income-stabilizing effect: remittances enhance borrowers’ capacity to repay (stable transfers can act as collateral), reduce income volatility, and improve banks’ ability to screen borrowers.
- Institutional and bank-specific factors affect credit quality: diversification, capitalization, bank size, public ownership, supervisory framework, and efficiency.
- Non-linearity: growth impact of remittances is maximized with good institutions; self-reinforcing mechanism implies stronger remittance effects in countries with higher institutional quality and financial development.
- Caution: increased income could also spur excess risk-taking by banks; credit risk may be limited if new loans are extended mainly to remittance receivers.

### III. Empirical framework and data
- Two model specifications estimated:
  - Linear effect: NPLi,t = θRi,t + φσi,t + X'i,tβ + ui + εi,t
    - NPLi,t: NPL ratio; Ri,t: remittances-to-GDP; σi,t: output growth volatility (five-year rolling standard deviation of real GDP growth per capita); X: traditional determinants including real GDP growth, GDP per capita, exports-to-GDP, credit boom dummy; governance index (PCA composite scaled 0–100); ui: country fixed effects.
  - Non-linear/stabilizing channel: NPLi,t = θ1 + θ2 σi,t × Ri,t + φσi,t + X'i,tβ + ui + εi,t (interaction between remittances and output volatility).
- Estimation methods:
  - OLS with country fixed effects.
  - Fractional logit (quasi-likelihood) to account for bounded dependent variable between zero and one and values of zero or one.
- Sample and data:
  - Unbalanced panel of 141 low- and middle-income countries over 2000–2011.
  - Remittances definition: World Bank aggregate (includes workers’ remittances and compensation of employees when not separable).
  - Credit boom defined following Mendoza and Terrones (2008): cyclical component of credit > 1.75 × its standard deviation (HP filter used).

### IV. Results — Baseline estimates and robustness
- Traditional determinants (Table 1):
  - Output growth volatility: positive and significant (OLS-FE 0.439; Fractional Logit 0.335).
  - Credit boom: positive and significant (OLS-FE 0.010; Fractional Logit 0.013).
  - Ln GDP per capita: negative and significant (OLS-FE -0.187; Fractional Logit -0.194).
  - Real GDP growth: negative and significant (OLS-FE -0.317; Fractional Logit -0.327).
  - Exports-to-GDP and Governance index: not significant in baseline FE regressions.
  - Sample stats: Observations 638; Number of countries 71; R2 0.32 (OLS-FE) and 0.67 (Fractional Logit).
- Linear remittances effects (Table 2):
  - Remittances-to-GDP coefficients range from -0.517 to -0.25 across specifications and methods (specific reported values: -0.517, -0.438, -0.515, -0.364, -0.25, -0.354).
  - Coefficients are negative and statistically significant (robust across inclusion/exclusion of credit boom and across OLS-FE and fractional logit).
  - Quantification: an increase in remittance-to-GDP by one half standard-deviation in the sample (4 percent of GDP) is associated with a drop in the NPL ratio of about 1–2 percentage points, ceteris paribus.
  - Controls maintain significance: output growth volatility, credit boom, Ln GDP per capita, and real GDP growth retain expected signs and significance.
  - Sample stats for Table 2: Observations 624; Number of countries 70; R2 ranges 0.32–0.35 (OLS-FE) and 0.67–0.68 (Fractional Logit).
- Non-linearity / stabilizing channel (Table 3):
  - Remittances-to-GDP additive term: OLS-FE -0.413; Fractional Logit -0.239 (both significant).
  - Output growth volatility: OLS-FE 0.714; Fractional Logit 0.634 (both significant).
  - Interaction Remittances-to-GDP × Output growth volatility: OLS-FE -1.911; Fractional Logit -1.715 (both significant), indicating remittances dampen the adverse effect of volatility on NPLs.
  - Joint significance of non-linearity: P-value 0.0002 (OLS-FE) and 0.0000 (Fractional Logit).
  - Reported remittances threshold (in percent of GDP): 35.5 (Table 3, OLS-FE reported value).
  - Sample stats for Table 3: Observations 624; Number of countries 70; R2 0.36 (OLS-FE) and 0.68 (Fractional Logit).

### V. Key empirical findings and implications
- Core empirical findings (bulleted):
  - Remittances are robustly negatively associated with NPL ratios in the sample of developing countries (point estimates for remittances-to-GDP range from -0.25 to -0.517).
  - Higher output growth volatility is positively associated with larger NPL ratios (coefficients reported: 0.335, 0.439, 0.481, 0.714, 0.634 across tables).
  - Credit booms are associated with higher NPLs (coefficients around 0.008–0.013).
  - Interaction results: remittances reduce the marginal impact of macroeconomic volatility on NPLs (interaction coefficients -1.911 and -1.715).
  - Non-linearities are econometrically significant (joint P-values 0.0002 and 0.0000).
- Interpretation:
  - Income-stabilizing effects of remittances appear to dominate the risk-inducing effects in the sample, implying remittances enhance borrowers’ capacity to repay and act as shock absorbers.
  - Countries with high macroeconomic volatility benefit more from remittances in terms of banking stability.

### VI. Data and descriptive statistics highlighted
- Time frame: 2000–2011.
- Sample: 141 developing countries; key estimation samples report 70–71 countries and 624–638 observations depending on specification.
- Remittances trend: average remittances to developing countries rose from $700 million (1990s average) to $2.8 billion in 2010 (text statement).
- Regional average NPLs (2000–11 sample averages, reported in Figure 4):
  - MENA: 13.9 percent
  - SAS: 12.8 percent
  - SSA: 11.0 percent
  - EAP: 10.9 percent
  - ECA: 9.0 percent
  - LAC: 5.2 percent
- Top recipients in sample by region (2007–11 average, percent of GDP) include:
  - SSA: Lesotho (35.5), Togo (10.5), Senegal (10.5), Cape Verde (9), Gambia (8.7)
  - EAP: Tonga (23.6), Samoa (20.2), Philippines (10.9), Vietnam (7.4), Fiji (4.6)
  - SAS: Nepal (21), Bangladesh (10.9), Sri Lanka (8), Pakistan (5), India (3.5)
  - MENA: Lebanon (21.2), Jordan (15.7), Morocco (7.5), Egypt (5.4), Yemen (5.1)
  - LAC: Haiti (21.3), Honduras (18.8), El Salvador (16.9), Jamaica (15.4), Guyana (15.1)

*Source: _wp14144 - References (IMF working paper content, 2000–2011 sample) .*

### 17. The results presented so far ignored an important issue that would significantly bias

### 17. The results presented so far ignored an important issue that would significantly bias

### Endogeneity concerns and sources of bias
- Endogeneity of remittances may arise from:
  - Reverse causality: countercyclicality of remittances with respect to financial shocks can be exacerbated in a context of high risks of default, as migrants insure those left behind by remitting money to help avoid repayment defaults.
  - Redundancy effect: balance of payments data capture primarily flows intermediated by banks, implying the sample relates to countries with a certain level of financial development, potentially interacting with NPL dynamics.
  - Omitted variable bias: unobserved and time-varying factors may be simultaneously correlated with the size of NPLs and remittance inflows (for example, migrants transferring norms and values that affect incentives to repay loans or the willingness to demand a loan for a well-structured project).
- These factors likely bias estimates and require an instrumental variable strategy.

### Instrumental variables identification strategy
- Primary instrument: real per capita GDP growth in remittance-sending countries, computed as the weighted average of GDP growth rates in all potential migrant destination countries with weights being the bilateral migration shares between the migrant-sending country and the migrant destination countries (based on the World Bank Bilateral Migration Database).
- Complementary instrument: two-year lag of the remittance variable (remittances-to-GDP in t-2).
- Identification assumptions:
  - Better economic conditions in migrant-destination countries are associated with larger remittances flowing into migrant-sending countries, conditional on country fixed-effects, the economic conditions in the remittance-receiving country (real per capita growth rate), and other sources of economic globalization (trade openness).
- Diagnostics performed:
  - Test of the strength of instruments in the first-stage equation.
  - Tests of the orthogonality of the instruments (Hansen over-identification test).

### Instrumental Variable results — Linear effect (Table 4)
- Sample and specification:
  - Dependent Variable: NPL Ratio
  - Period: 2000–11
  - Observations: 560
  - Number of countries: 65
  - Country fixed-effects: Yes
- Second-stage estimates (coefficient (robust standard error); significance):
  - Remittances-to-GDP: -0.636 (0.183)*** 
  - Output growth volatility: 0.380 (0.130)*** 
  - Credit boom: 0.006 (0.004) 
  - Ln GDP per capita: -0.199 (0.016)*** 
  - Real GDP growth: -0.295 (0.061)*** 
  - Exports-to-GDP: -0.002 (0.049) 
  - Governance index: -0.003 (0.011)
- First-stage indicators:
  - Growth in migrant host countries: 0.09 (0.041)** 
  - Remittances-to-GDP (in t-2): 0.493 (0.060)*** 
  - F-statistic of remittances-to-GDP Instrumental equation: 54.2
  - Hansen over-identification test (p-value): 0.57
- Interpretation:
  - Instrumental variable estimates confirm a negative and robust association between remittance inflows and the decline in NPL ratios.
  - Point estimates from IV are in absolute terms higher than OLS counterparts, suggesting the direction of endogeneity bias was positive.
  - IV helps rule out positive reverse causality running from higher risk of default to higher remittance inflows.
  - Instruments are strongly correlated with remittances, have expected signs, and perform well.

### Instrumental Variable results — Non-linearity and interaction with volatility (Table 5)
- Sample and specification:
  - Dependent Variable: NPL Ratio
  - Period: 2000–11
  - Observations: 560
  - Number of countries: 65
  - Country fixed-effects: Yes (both first-stage and second-stage)
- Second-stage estimates (coefficient (robust standard error); significance):
  - Remittances-to-GDP: -0.485 (0.172)*** 
  - Remittances-to-GDP * output growth volatility: -1.301 (0.728)* 
  - Output growth volatility: 0.539 (0.197)*** 
  - Credit boom: 0.005 (0.004) 
  - Ln GDP per capita: -0.201 (0.015)*** 
  - Real GDP growth: -0.285 (0.063)*** 
  - Exports-to-GDP: -0.008 (0.048) 
  - Governance index: -0.003 (0.011)
- First-stage (two equations reported: Remittances-to-GDP; Remittances-to-GDP*Output Growth Volatility):
  - Growth in migrant host countries: 0.121 (0.036)*** ; 0.002 (0.001)
  - Remittances-to-GDP (in t-2): 0.514 (0.069)*** ; -0.005 (0.003)* 
  - Growth in migrant host countries*output growth volatility: -1.006 (1.012) ; -0.034 (0.65)
  - Remittances-to-GDP (in t-2)*output growth volatility: -0.196 (0.157) ; 0.568 (0.008)*** 
  - F-statistic of instrumentation equation: 31.6 ; 2624.0
  - Hansen over-identification test (p-value): 0.66
  - Joint significance of the non-linearity: P-val 0.0001
- Interpretation:
  - Instrumentation procedure does not alter previous findings.
  - Strong and robust effect of remittances on NPLs is reinforced in countries characterized by high macroeconomic volatility.
  - Both linear and nonlinear components of remittances were instrumented by entering instruments linearly and in interaction with output growth volatility in first-stage regressions.

### Robustness checks and alternative measures
- Alternative remittance measure:
  - Remittances-to-deposits used as an alternative measure; results in the Appendix confirm the robust and negative impact of remittances on the size of NPLs.
  - Table A3 coefficient estimates for Remittances-to-deposit across specifications:
    - (1) -0.020 (0.011)* 
    - (2) -0.020 (0.007)*** 
    - (3) -0.019 (0.007)*** 
    - (4) -0.019 (0.007)*** 
    - (5) -0.017 (0.008)** 
  - Additional Appendix statistics (Table A3):
    - # Countries: 59
    - R2: 0.03, 0.37, 0.38, 0.38, 0.40 across columns (1)–(5)
    - N: 475, 475, 473, 473, 473
- Sample bias sensitivity:
  - Dividing the country sample into different income groups (low income, lower middle and upper middle income categories) broadly confirms previous results (available upon request).
- Descriptive statistics (Table A1):
  - NPL ratio: Obs 624; Mean 0.09; Std. Dev. 0.08; Min 0.00; Max 0.57
  - Remittances-to-GDP: Obs 624; Mean 0.05; Std. Dev. 0.07; Min 0.00; Max 0.51
  - Output growth volatility: Obs 624; Mean 0.03; Std. Dev. 0.03; Min 0.00; Max 0.34
  - Credit boom: Obs 624; Mean 0.43; Std. Dev. 0.50; Min 0.00; Max 1.00
  - ln GDP per capita: Obs 624; Mean 7.48; Std. Dev. 0.96; Min 5.03; Max 9.36
  - Real GDP growth: Obs 624; Mean 0.05; Std. Dev. 0.04; Min -0.18; Max 0.27
  - Exports-to-GDP: Obs 624; Mean 0.37; Std. Dev. 0.18; Min 0.07; Max 1.20
  - Governance index: Obs 624; Mean 5.91; Std. Dev. 1.35; Min 2.79; Max 10.00

### Conclusion (V. CONCLUSION)
- Findings:
  - Results confirm the role of traditional determinants of NPLs: macroeconomic factors together with banks’ specific and institutional factors.
  - Remittances can reduce NPLs by enhancing and stabilizing borrowers’ capacity to repay.
  - Remittances can be expected to pave the way for a sound development of a formal financial sector, critical for sustaining high economic growth.

*IMF Working Paper content (excerpt provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14144.pdf_