## 7. Impulse Response Functions, Alternative Specification

## Source details

**Canonical URL:** [7. Impulse Response Functions, Alternative Specification](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1372.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1372.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1372.pdf.json)

---

### Introduction and key context
- High and rising levels of non-performing loans (NPLs) in CESEE exert strong pressure on banks’ balance sheets and may adversely affect banks’ lending operations.
- NPLs in the region increased to an average of 11 percent (end-2011) from just above 3 percent in 2007.
- The surge in NPLs is widespread across banks and evident in both retail and corporate segments.
- Policymakers have placed the resolution of the NPLs problem as a priority.

### Objectives of the study
- Evaluate the determinants of non-performing loans in CESEE using bank-level data and macroeconomic indicators over 1998–2011.
- Examine how the relative importance of bank-level vs. macroeconomic factors changed since the onset of the financial crisis.
- Evaluate feedback effects from the banking sector to the real economy using a panel vector auto-regression (VAR) including five endogenous variables (NPLs, real GDP growth, unemployment rate, change in credit-to-GDP ratio, and inflation).

### Data and empirical approach
- Panel data: individual banks’ balance sheets from Bankscope; macroeconomic indicators from Haver and WEO.
- Annual frequency for 1998–2011; covers the ten largest banks in each of 16 countries (Bosnia and Herzegovina; Bulgaria; Croatia; Czech Rep.; Estonia; Hungary; Latvia; Lithuania; Macedonia; Poland; Romania; Russia; Serbia; Slovak Rep.; Slovenia; Ukraine).
- Dataset coverage: above 60 percent of the banking sector’s assets in most countries (Table 1).
- Baseline explanatory variables:
  - Bank-level: equity-to-assets ratio, return on equity (RoE), Loans-to-Assets ratio, Loans growth rate (D_Loans).
  - Country-specific: inflation (Inf), change in exchange rate vis-à-vis the euro (Ex_rate), change in unemployment rate (D_unemp).
  - Global: Euro zone’s GDP growth (D_rgdp_euro), global risk aversion (VIX).
- Bankscope reports “impaired loans” treated as NPLs in this analysis.
- Sample sizes and distributions:
  - NPLs: 976 bank-level observations; median NPL climbed from just below 3 percent (pre-crisis) to about 11 percent in 2011.
  - Country-level panel VAR: 206 annual observations of NPLs for 16 CESEE economies, 1998–2011.

### Econometric specifications
- Dynamic panel regression using:
  - Fixed effects model.
  - Difference GMM (Arellano and Bond, 1991).
  - System GMM (Arellano and Bover, 1995; Blundell and Bond, 1998).
- Dependent variable: logit transformation of the NPLs ratio to map NPLs onto (-∞, ∞).
- Three caveats highlighted:
  - Potential inconsistencies in NPL definitions across countries and supervisors.
  - Restructured and “ever-greened” loans may mask true NPLs in some countries.
  - Composition of outstanding loans (currency, corporate vs. retail, housing vs. consumption) not controlled due to data limitations.

### Main empirical findings — determinants of NPLs
- NPLs affected by both macroeconomic and bank-level factors; bank-level factors have significant but relatively low overall explanatory power.
- Persistence of NPLs: coefficient on lagged NPLs ranges between 0.6 to 0.93, indicating prolonged effects of shocks to NPLs.
- Macroeconomic determinants (direction and significance as reported):
  - Change in unemployment (D_unemp): positive effect on NPLs; e.g., coefficients include 0.040**, 0.010, 0.026, -0.002, 0.049*, 0.039*** across specifications.
  - Inflation (Inf (-1)): higher inflation increases NPLs in several specifications (coefficients reported include 0.006, 0.026*, 0.030*, 0.038**, 0.012***).
  - Exchange rate depreciation (Ex_rate): depreciation increases NPLs; coefficients include 0.003, 0.006, 0.012*, 0.009**, 0.005, 0.009** (An increase in exchange rate indicates depreciation).
  - Euro area real GDP growth (D_rgdp_euro (-1)): higher Euro area growth reduces NPLs; coefficients include -0.053*, -0.034*, -0.038, -0.028**, -0.030*, -0.017***.
  - Global risk aversion (Vix): higher VIX increases NPLs; coefficients include 0.028*, 0.020*, 0.022, 0.014**, 0.024*, 0.022*.
- Bank-level determinants:
  - Equity-to-Assets (-1): higher equity-to-assets reduces NPLs (coefficients -0.040*, -0.061*, -0.044**).
  - Return on Equity (RoE (-1)): higher RoE reduces NPLs (coefficients -0.008*, -0.000, -0.005**).
  - Loans-to-Assets (-1): higher loans-to-assets associated with higher NPLs in some specifications (coefficients 0.017*, 0.032*).
  - D_loans(-2): lagged loan growth raises NPLs in some specifications (coefficients 0.000, 0.001, 0.002***).
- Goodness-of-fit and diagnostics (selected reported values):
  - R-squared (within): 0.601 and 0.620 reported for fixed effects specifications.
  - R-squared (between): 0.830 and 0.693 reported.
  - Hansen test p-values: 0.624 and 0.206.
  - Arellano-Bond AR(1) test p-values: 0.000 and 0.000; AR(2) test p-values: 0.289 and 0.433.

### Panel VAR methodology for feedback effects
- Variables in panel VAR (Y_it): NPLs (non-performing loans-to-total loans of overall banking system), Δcredit (change in credit-to-GDP ratio), Δrgdp (real GDP growth), unemp (unemployment rate), Δcpi (inflation).
- Estimation approach: panel VAR with forward mean-differencing (Helmert procedure) and GMM instruments; shocks orthogonalized via Cholesky decomposition.
- Ordering used (for Cholesky): NPLs first, then Δcredit, unemp, Δrgdp, Δcpi (presumption that GDP, unemployment, and inflation affect NPLs only with a lag while NPLs contemporaneously affect economic activity mainly through credit). Results qualitatively robust to alternative orderings.

### Panel VAR results — impulse response findings
- Response of NPLs to shocks in other variables (baseline specification):
  - A one percentage point increase in credit-to-GDP ratio leads to a cumulative decline of 0.7 percentage point in NPLs (in the subsequent year).
  - A one percentage point increase in real GDP leads to a cumulative decline of 0.6 percentage point in NPLs (in the subsequent year).
  - A one percentage point increase in inflation leads to a cumulative increase of 0.4 percentage point in NPLs (in the subsequent year).
  - In this specification a shock to unemployment was not found to have a significant impact on NPLs; in an alternative specification with first differences, unemployment impact was significant.
- Impact of a shock to NPLs:
  - A one percentage point increase in NPLs results in a cumulative decline in credit, inflation, and real GDP growth, while contributing to higher unemployment (described as “negative and significant” for credit and real GDP).
  - A positive shock to GDP growth and credit contribute to the reduction of NPLs while higher inflation leads to higher NPLs.
  - A positive shock (increase) to NPLs ratio leads to a contraction of credit-to-GDP ratio and real GDP and to a higher unemployment rate; as a result of weaker economic activity, CPI inflation also declines.

### Robustness and sub-sample analysis
- Sample split: pre-crisis (1998–2007) vs. post-crisis (2008–2011).
- Findings from sub-samples:
  - Inflation and the change in unemployment had larger impacts on NPLs during the pre-crisis period.
  - In the post-crisis period, exchange rate contribution to NPLs was more prominent; inflation was not found significant.
  - Bank-level factors matter in both periods, though coefficient significance depends on estimation technique.
- Stationarity tests:
  - Panel unit-root tests (Fisher-ADF and Fisher-PP) indicate the five endogenous variables are stationary I(0) in at least one test; the null of unit root in all panels can be rejected with 95 percent confidence or higher in those tests.

### Key quantitative impacts of NPL shocks (baseline and alternative specifications)
- Baseline specification:
  - A shock produces a reduction of 1.7 percentage points in credit-to-GDP ratio.
  - A cumulative increase of 0.5 in unemployment (over three years).
  - A cumulative contraction of about one percentage point in real GDP (over two years).
  - A cumulative decline in inflation of 0.6 percentage points (over three year horizon).
  - Note: The impact of NPL on credit and GDP growth was found to be rather large compared to previous findings; point estimates have relatively wide confidence intervals and are sensitive to sample composition.
- Alternative specification (changes / D_ variables):
  - An increase of one percentage point in unemployment results in an increase of NPLs in the subsequent year by 0.3 percentage point.
  - An increase of one percentage point in inflation results in an increase of NPLs in the subsequent year by 0.5 percentage point.
  - An increase of one percentage point in real GDP leads to a decline of 0.8 percentage points in NPLs (including through an expansion of credit) in the subsequent year.
  - A one percentage point increase in NPLs results in a cumulative decline of 1.5 percentage point in credit-to-GDP ratio.
  - A cumulative increase of nearly 0.5 percentage points in unemployment (over a three-year horizon).
  - A cumulative contraction of 1.3 percentage points in real GDP over two years.
  - A cumulative decline in inflation of 0.4 percentage points over three years.

### Variance decomposition and explanatory power
- In a 5-year horizon, NPLs explains about 10 percent of the forecast error of the rest of the endogenous variables.
- The change in credit has the most information regarding the variation of NPLs, followed by inflation and real GDP growth.
- In the alternative specification:
  - The explanatory power of D_NPL regarding the variation of other variables is somewhat higher (10-20 percent).
  - The explanatory power of other endogenous variables with regards to the variation of D_NPL is on average lower.
- Selected figures from Table 6 (percent of variation in the row variable explained by column variable, 5-year horizon):
  - Baseline: NPL row — NPL: 0.573; D_CREDIT: 0.212; UNP: 0.009; D_RGDP: 0.109; D_CPI: 0.097.
  - Alternative estimation: D_NPL row — D_NPL: 0.689; D_CREDIT: 0.152; D_UNP: 0.015; D_RGDP: 0.072; D_CPI: 0.073.

### Determinants of NPLs — succinct summary
- Macroeconomic drivers:
  - NPLs tend to increase when unemployment rises, exchange rate depreciates, and inflation is high.
  - Beyond country-specific effects, euro area GDP growth and global risk aversion have direct impacts on banks’ asset quality.
- Bank-level drivers:
  - Higher quality of bank management (measured by previous period’s profitability) leads to lower NPLs.
  - Moral hazard incentives, such as low equity, tend to worsen NPLs.
  - Excessive risk-taking (measured by loans-to-assets ratio and the growth rate of bank’s loans) contributes to higher NPLs in subsequent periods.
  - These bank-level effects were significant during both the pre-crisis and post-crisis periods.

### Conclusions and policy implications
- Rapid deterioration in asset quality in many CESEE banks led to substantial losses and reduction of capital buffers, limiting lending and affecting economic activity.
- There are significant feedback effects between the banking system and the real economy: increases in NPLs materially affect credit (as a share of GDP), real GDP growth, unemployment, and inflation in subsequent periods.
- Policy recommendations:
  - Strengthen supervision to prevent sharp buildup of NPLs, including ensuring banks avoid excessive lending and maintain high credit standards.
  - Limit foreign currency lending to un-hedge borrowers.
  - Undertake a swift, but orderly, clean-up of banks’ portfolios given the burden high levels of NPLs pose on the economy.
  - Prefer bank-led, collective and cooperative resolution of NPLs that benefits debtors and creditors, but policymakers can be proactive by removing tax, legal, and regulatory impediments to help banks accelerate the cleanup process in a non-disruptive manner while taking into account banks’ ability to absorb losses.

*Source: _wp1372 - 7. Impulse Response Functions, Alternative Specification (sections and tables as provided in the supplied content).*

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

### _wp1372 - References

### Tables
- 1. Coverage of Dataset as a Percent of the Banking system Total Assets .................................8
- 2. Summary Statistics, Bank-Level Indicators, 1998-2011 .......................................................9
- 3. Summary Statistics, Country-Specific and Global Indicators, 1998-2011 ............................9
- 4. NPLs: Macroeconomic and Bank-Level Determinants, 1998-2011 ....................................11
- 5. Correlation Matrix ...............................................................................................................15
- 6. Variance Decomposition ......................................................................................................19

### Figures
- 1. Non-Performing Loans in CESEE Economies ......................................................................3
- 2. Real GDP Growth and the Change in NPLS in CESEE Economies, 200-2011 ....................3
- 3. NPLs at Bank-Level ...............................................................................................................9
- 4. Distribution of the Level and the Change of Non-Performing Loans, 1998-2011 ..............15
- 5. Correlation between the Change in NPLs, Unemployment and GDP Growth ....................15
- 6. Impulse Response Functions................................................................................................17

*Source: _wp1372 - References.*

### 7. Impulse Response Functions, Alternative Specification .....................................................18

### 7. Impulse Response Functions, Alternative Specification

### Introduction and key context
- High and rising levels of non-performing loans (NPLs) in CESEE exert strong pressure on banks’ balance sheets and may adversely affect banks’ lending operations.
- NPLs in the region increased to an average of 11 percent (end-2011) from just above 3 percent in 2007.
- The surge in NPLs is widespread across banks and evident in both retail and corporate segments.
- Policymakers have placed the resolution of the NPLs problem as a priority.

### Objectives of the study
- Evaluate the determinants of non-performing loans in CESEE using bank-level data and macroeconomic indicators over 1998–2011.
- Examine how the relative importance of bank-level vs. macroeconomic factors changed since the onset of the financial crisis.
- Evaluate feedback effects from the banking sector to the real economy using a panel vector auto-regression (VAR) including five endogenous variables (NPLs, real GDP growth, unemployment rate, change in credit-to-GDP ratio, and inflation).

### Data and empirical approach
- Panel data: individual banks’ balance sheets from Bankscope; macroeconomic indicators from Haver and WEO.
- Annual frequency for 1998–2011; covers the ten largest banks in each of 16 countries (Bosnia and Herzegovina; Bulgaria; Croatia; Czech Rep.; Estonia; Hungary; Latvia; Lithuania; Macedonia; Poland; Romania; Russia; Serbia; Slovak Rep.; Slovenia; Ukraine).
- Dataset coverage: above 60 percent of the banking sector’s assets in most countries (Table 1).
- Baseline explanatory variables:
  - Bank-level: equity-to-assets ratio, return on equity (RoE), Loans-to-Assets ratio, Loans growth rate (D_Loans).
  - Country-specific: inflation (Inf), change in exchange rate vis-à-vis the euro (Ex_rate), change in unemployment rate (D_unemp).
  - Global: Euro zone’s GDP growth (D_rgdp_euro), global risk aversion (VIX).
- Bankscope reports “impaired loans” treated as NPLs in this analysis.
- Sample sizes and distributions:
  - NPLs: 976 bank-level observations; median NPL climbed from just below 3 percent (pre-crisis) to about 11 percent in 2011.
  - Country-level panel VAR: 206 annual observations of NPLs for 16 CESEE economies, 1998–2011.

### Econometric specifications
- Dynamic panel regression using:
  - Fixed effects model.
  - Difference GMM (Arellano and Bond, 1991).
  - System GMM (Arellano and Bover, 1995; Blundell and Bond, 1998).
- Dependent variable: logit transformation of the NPLs ratio to map NPLs onto (-∞, ∞).
- Three caveats highlighted:
  - Potential inconsistencies in NPL definitions across countries and supervisors.
  - Restructured and “ever-greened” loans may mask true NPLs in some countries.
  - Composition of outstanding loans (currency, corporate vs. retail, housing vs. consumption) not controlled due to data limitations.

### Main empirical findings — determinants of NPLs
- NPLs affected by both macroeconomic and bank-level factors; bank-level factors have significant but relatively low overall explanatory power.
- Persistence of NPLs: coefficient on lagged NPLs ranges between 0.6 to 0.93, indicating prolonged effects of shocks to NPLs.
- Macroeconomic determinants (direction and significance as reported):
  - Change in unemployment (D_unemp): positive effect on NPLs; e.g., coefficients include 0.040**, 0.010, 0.026, -0.002, 0.049*, 0.039*** across specifications (see Table 4).
  - Inflation (Inf (-1)): higher inflation increases NPLs in several specifications (coefficients reported include 0.006, 0.026*, 0.030*, 0.038**, 0.012***).
  - Exchange rate depreciation (Ex_rate): depreciation increases NPLs; coefficients include 0.003, 0.006, 0.012*, 0.009**, 0.005, 0.009** (An increase in exchange rate indicates depreciation).
  - Euro area real GDP growth (D_rgdp_euro (-1)): higher Euro area growth reduces NPLs; coefficients include -0.053*, -0.034*, -0.038, -0.028**, -0.030*, -0.017***.
  - Global risk aversion (Vix): higher VIX increases NPLs; coefficients include 0.028*, 0.020*, 0.022, 0.014**, 0.024*, 0.022*.
- Bank-level determinants:
  - Equity-to-Assets (-1): higher equity-to-assets reduces NPLs (coefficients -0.040*, -0.061*, -0.044**).
  - Return on Equity (RoE (-1)): higher RoE reduces NPLs (coefficients -0.008*, -0.000, -0.005**).
  - Loans-to-Assets (-1): higher loans-to-assets associated with higher NPLs in some specifications (coefficients 0.017*, 0.032*).
  - D_loans(-2): lagged loan growth raises NPLs in some specifications (coefficients 0.000, 0.001, 0.002***).
- Goodness-of-fit and diagnostics (selected reported values):
  - R-squared (within): 0.601 and 0.620 reported for fixed effects specifications.
  - R-squared (between): 0.830 and 0.693 reported.
  - Hansen test p-values: 0.624 and 0.206 (suggesting instruments uncorrelated with residuals).
  - Arellano-Bond AR(1) test p-values: 0.000 and 0.000; AR(2) test p-values: 0.289 and 0.433.

### Panel VAR methodology for feedback effects
- Variables in panel VAR (Y_it): NPLs (non-performing loans-to-total loans of overall banking system), Δcredit (change in credit-to-GDP ratio), Δrgdp (real GDP growth), unemp (unemployment rate), Δcpi (inflation).
- Estimation approach: panel VAR with forward mean-differencing (Helmert procedure) and GMM instruments; shocks orthogonalized via Cholesky decomposition.
- Ordering used (for Cholesky): NPLs first, then Δcredit, unemp, Δrgdp, Δcpi (presumption that GDP, unemployment, and inflation affect NPLs only with a lag while NPLs contemporaneously affect economic activity mainly through credit). Results qualitatively robust to alternative orderings.

### Panel VAR results — impulse response findings
- Response of NPLs to shocks in other variables:
  - A one percentage point increase in credit-to-GDP ratio leads to a cumulative decline of 0.7 percentage point in NPLs (in the subsequent year).
  - A one percentage point increase in real GDP leads to a cumulative decline of 0.6 percentage point in NPLs (in the subsequent year).
  - A one percentage point increase in inflation leads to a cumulative increase of 0.4 percentage point in NPLs (in the subsequent year).
  - In this specification a shock to unemployment was not found to have a significant impact on NPLs; in an alternative specification with first differences, unemployment impact was significant.
- Impact of a shock to NPLs:
  - A one percentage point increase in NPLs results in a cumulative decline in credit, inflation, and real GDP growth, while contributing to higher unemployment (exact magnitudes for these cumulative declines are described qualitatively in the text; the specific numeric cumulative declines for credit, inflation, and real GDP are stated as “negative and significant”).
  - The impulse response functions reveal that a positive shock to GDP growth and credit contribute to the reduction of NPLs while higher inflation leads to higher NPLs.
  - A positive shock (increase) to NPLs ratio leads to a contraction of credit-to-GDP ratio and real GDP and to a higher unemployment rate; as a result of weaker economic activity, CPI inflation also declines.

### Robustness and sub-sample analysis
- Sample split: pre-crisis (1998–2007) vs. post-crisis (2008–2011).
- Findings from sub-samples:
  - Inflation and the change in unemployment had larger impacts on NPLs during the pre-crisis period.
  - In the post-crisis period, exchange rate contribution to NPLs was more prominent; inflation was not found significant.
  - Bank-level factors matter in both periods, though coefficient significance depends on estimation technique.
- Stationarity tests:
  - Panel unit-root tests (Fisher-ADF and Fisher-PP) indicate the five endogenous variables are stationary I(0) in at least one test; the null of unit root in all panels can be rejected with 95 percent confidence or higher in those tests.

### Interpretations and policy-relevant implications (as presented)
- Macroeconomic stabilization supports bank asset quality: higher GDP growth and credit growth reduce NPLs; higher unemployment, inflation, exchange-rate depreciation, and global risk aversion increase NPLs.
- Bank prudential and supervisory stance matters: higher capitalization (equity-to-assets) and better profitability (RoE) are associated with lower NPLs; excessive lending and rapid loan growth raise future NPLs.
- Feedback loop: rising NPLs reduce credit supply and economic activity and raise unemployment, which can further worsen NPLs—highlighting a macro-financial amplification channel.
- Given these dynamics, prioritizing NPL resolution and policies that bolster macroeconomic recovery and bank capitalization are presented as critical in mitigating persistent bank-sector stress and supporting a sustained recovery.

*Source: _wp1372 - 7. Impulse Response Functions, Alternative Specification (sections and tables as provided in the supplied content).*

### 1.7 percentage points in credit-to-GDP ratio and a cumulative increase of 0.5 in

### _wp1372 - 1.7 percentage points in credit-to-GDP ratio and a cumulative increase of 0.5 in

### Key quantitative impacts of NPL shocks (baseline specification)
- A shock produces a reduction of 1.7 percentage points in credit-to-GDP ratio.
- A cumulative increase of 0.5 in unemployment (over three years).
- A cumulative contraction of about one percentage point in real GDP (over two years).
- A cumulative decline in inflation of 0.6 percentage points (over three year horizon).
- Note: The impact of NPL on credit and GDP growth was found to be rather large compared to previous findings; point estimates have relatively wide confidence intervals and are sensitive to sample composition.

### Responses and transmission channels (alternative specification and robustness)
- Replacing levels of NPL and unemployment with their changes yields broadly similar dynamics with some magnitude differences.
- Response of NPLs to shocks in other variables:
  - An increase of one percentage point in unemployment results in an increase of NPLs in the subsequent year by 0.3 percentage point.
  - An increase of one percentage point in inflation results in an increase of NPLs in the subsequent year by 0.5 percentage point.
  - An increase of one percentage point in real GDP leads to a decline of 0.8 percentage points in NPLs (including through an expansion of credit) in the subsequent year.
- Impact of a shock to NPLs (alternative specification):
  - A one percentage point increase in NPLs results in a cumulative decline of 1.5 percentage point in credit-to-GDP ratio.
  - A cumulative increase of nearly 0.5 percentage points in unemployment (over a three-year horizon).
  - A cumulative contraction of 1.3 percentage points in real GDP over two years.
  - A cumulative decline in inflation of 0.4 percentage points over three years.

### Variance decomposition and explanatory power
- In a 5-year horizon, NPLs explains about 10 percent of the forecast error of the rest of the endogenous variables.
- The change in credit has the most information regarding the variation of NPLs, followed by inflation and real GDP growth.
- In the alternative specification:
  - The explanatory power of D_NPL regarding the variation of other variables is somewhat higher (10-20 percent).
  - The explanatory power of other endogenous variables with regards to the variation of D_NPL is on average lower.
- Selected figures from Table 6 (percent of variation in the row variable explained by column variable, 5-year horizon):
  - Baseline: NPL row — NPL: 0.573; D_CREDIT: 0.212; UNP: 0.009; D_RGDP: 0.109; D_CPI: 0.097.
  - Alternative estimation: D_NPL row — D_NPL: 0.689; D_CREDIT: 0.152; D_UNP: 0.015; D_RGDP: 0.072; D_CPI: 0.073.

### Determinants of NPLs (macroeconomic and bank-level findings)
- Macroeconomic drivers:
  - NPLs tend to increase when unemployment rises, exchange rate depreciates, and inflation is high.
  - Beyond country-specific effects, euro area GDP growth and global risk aversion have direct impacts on banks’ asset quality.
- Bank-level drivers:
  - Higher quality of bank management (measured by previous period’s profitability) leads to lower NPLs.
  - Moral hazard incentives, such as low equity, tend to worsen NPLs.
  - Excessive risk-taking (measured by loans-to-assets ratio and the growth rate of bank’s loans) contributes to higher NPLs in subsequent periods.
  - These bank-level effects were significant during both the pre-crisis and post-crisis periods.

### Conclusions and policy implications
- Rapid deterioration in asset quality in many CESEE banks led to substantial losses and reduction of capital buffers, limiting lending and affecting economic activity.
- There are significant feedback effects between the banking system and the real economy: increases in NPLs materially affect credit (as a share of GDP), real GDP growth, unemployment, and inflation in subsequent periods.
- Policy recommendations (textual prescriptions from the paper):
  - Strengthen supervision to prevent sharp buildup of NPLs, including ensuring banks avoid excessive lending and maintain high credit standards.
  - Limit foreign currency lending to un-hedge borrowers.
  - Undertake a swift, but orderly, clean-up of banks’ portfolios given the burden high levels of NPLs pose on the economy.
  - Prefer bank-led, collective and cooperative resolution of NPLs that benefits debtors and creditors, but policymakers can be proactive by removing tax, legal, and regulatory impediments to help banks accelerate the cleanup process in a non-disruptive manner while taking into account banks’ ability to absorb losses.

*Source: IMF working paper content provided in the source PDF.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1372.pdf_