## _wp10224

## Source details

**Canonical URL:** [_wp10224](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10224.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10224.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10224.pdf.json)

---

### Introduction and methodology
- Focus: relationship between macroeconomic variables and NPLs (credit risk) in GCC banks using bank-level data from Bankscope.
- Sample/coverage: dynamic panel estimated over 1995–2008 on around 80 banks in the GCC region.
- Dependent variable: logit transformation of the NPLs ratio (log(NPLs/(1-NPLs))).
- Models estimated: OLS, fixed effects, Arellano-Bond difference GMM, System GMM (with collapsed instruments and forward orthogonalization).
- Macroeconomic panel VAR estimated to investigate feedback effects of bank performance on credit supply and growth.

### Main empirical summary
- NPL ratios worsen as economic growth becomes lower and interest rates increase.
- Larger banks and banks with lower expenses tend to have lower NPLs.
- High past credit growth can generate higher NPLs in the future.
- NPLs are very persistent; AR coefficient of logit(NPL) estimated between 0.6 and 0.9.
- Cumulative effect of macroeconomic shocks over a three year horizon is large.
- Policy implication highlighted: in the context of exchange rate pegs, a stronger focus on macroprudential regulation—particularly through capital and liquidity buffers and countercyclical provisioning norms—would help mitigate macroeconomic risks on the banking system and the feedback effect of credit risks on the economy.

### Stylized facts on GCC banking system
- Historical context:
  - GCC experienced significant increases in banking system credit between 2003 and 2008; favorable pre-crisis macro environment associated with lower NPLs.
  - NPL ratios in double-digits were not uncommon historically before the boom years.
- 2009 country NPL increases (reported):
  - Bahrain 3.9 percent, Kuwait 9.7 percent, Oman 2.8 percent, Qatar 1.7 percent, Saudi Arabia 3.3 percent, U.A.E. 4.6 percent.
- Summary statistics (Bankscope, 1995–2008, unweighted NPL ratio in 2008 and percentiles):
  - NPL ratio in 2008 (unweighted): Bahrain 2.9, Kuwait 5.5, Oman 1.2, Qatar 1.7, Saudi Arabia 1.4, UAE 2.4
  - NPL ratio (1995–2008) percentiles (by country median unless otherwise noted):
    - 25th percentile: Bahrain 2.1, Kuwait 3.4, Oman 3.5, Qatar 1.3, Saudi Arabia 1.7, UAE 1.7
    - Median: Bahrain 5.9, Kuwait 6.7, Oman 6.5, Qatar 3.1, Saudi Arabia 3.1, UAE 4.3
    - 75th percentile: Bahrain 14.8, Kuwait 12.5, Oman 11.3, Qatar 11.5, Saudi Arabia 7.3, UAE 9.6
- Heterogeneity: in downturns, NPLs increase much faster for banks with higher initial NPL levels (nonlinear amplification across banks).

### Determinants of credit risk — bank-level panel results (selected coefficients)
- Autoregressive term ln(NPL/(1-NPL)) lagged:
  - OLS 0.898***, FE 0.676***, Arellano-Bond 0.714***, System GMM (collapsed) 0.881***, System GMM (fwd. orth., collapsed) 0.865***.
- ln(equity) lagged:
  - OLS -0.0543**, FE -0.364***, Arellano-Bond -0.359***, System GMM -0.102, System GMM (fwd) -0.102.
- (expenses/avg. assets) lagged:
  - OLS 0.0483*, FE 0.0868*, Arellano-Bond 0.154**, System GMM 0.114**, System GMM (fwd) 0.106**.
- loans growth lagged:
  - OLS 0.104**, FE 0.0946*, Arellano-Bond 0.0993*, System GMM 0.137, System GMM (fwd) 0.145.
- non-oil GDP growth lagged:
  - OLS -1.948***, FE -0.974, Arellano-Bond -1.893**, System GMM -2.156*, System GMM (fwd) -2.090*.
- interest rate lagged:
  - OLS 0.0241*, FE 0.0535**, Arellano-Bond 0.0044, System GMM -0.0053, System GMM (fwd) 0.0001.
- VIX (global risk aversion) contemporaneous/lagged:
  - OLS 0.0131***, FE 0.0115***, Arellano-Bond 0.0119***, System GMM 0.0140***, System GMM (fwd) 0.0132***.
- Model diagnostics and sample:
  - Observations: 426 (reported columns).
  - Number of banks in sample: 79 (varies by specification).
  - No. of instruments in GMM: kept below 40; Hansen test p-values reported (for example 0.27, 0.48, 0.52); A-B AR(1) test p-values 0.00; A-B AR(2) test p-values 0.17–0.19.

### Interpretations and quantitative implications from panel estimates
- Persistence: estimated AR coefficient between 0.6 (FE biased downward) and 0.9 (OLS biased upward) implies slow response to shocks but potential cumulation of NPLs over time.
- Macroeconomic shocks:
  - A temporary decrease by 3 percentage points in non-oil GDP growth would increase NPLs by 0.3 to 1.1 percentage points, depending on initial NPL levels (non-linearity of logit model implies marginal effects vary across the distribution).
  - A 300 basis points increase in interest rates would have a broadly similar effect on NPLs.
- VIX significance: external financing conditions/global risk aversion matter for GCC credit risk.
- World trade growth and the Asian crisis dummy were not significant.
- Lagged credit growth is associated with higher future NPLs.
- Non-linearities: shocks have larger effects for banks with lower-quality loan portfolios; interaction terms tested were not significant—likely captured by logit non-linearity.

### Feedback from the banking system to the real economy (Panel VAR results)
- Purpose: estimate effect of increases in NPLs on credit growth and non-oil GDP growth using panel VAR for GCC banking-system NPLs.
- Data notes:
  - Banking-system NPL data available since late 1990s for countries (examples: Bahrain since 1999; Kuwait since 1996; Oman since 1997; Qatar since 1999; Saudi Arabia since 1997; U.A.E. since 1996).
  - Small and unbalanced panel; unit-root tests (Levin-Lin-Chu) results not robust to sample period.
- Stationarity/unit-root testing (Levin-Lin-Chu, selected):
  - NPLs/Total Loans 1999-2008: 12.7*** (level), NPLs/Total Loans 2000-2008: -0.6 (level), difference -4.6***.
  - Non-oil real GDP growth 1998-2008: -3.3*** (level), 1999-2008: -0.7 (level), difference -1.4*.
  - Credit growth 1998-2008: 1.9 (level), difference -3.2***.
  - Interest rate 1998-2008: -6.2*** (level).
  - Significance levels: * 10%; ** 5%; *** 1%.
- VAR specification and identification:
  - Two panel VARs estimated: one with NPL ratio in levels, one with NPL ratio in first differences.
  - Lag length restricted to 2 lags.
  - Variables demeaned using Helmert procedure.
  - Choleski identification ordering: interest rate first, credit growth (or non-oil real GDP growth) second, NPL ratio last.

### Key panel-VAR empirical findings (impulse responses and decomposition)
- Panel VAR suggests a strong albeit short-lived feedback effect of higher NPLs on non-oil growth, reported as a semi-elasticity of around 0.4.
- In VAR with NPLs in levels and ordering NPL last:
  - Higher interest rates increase NPLs (effect not always significant).
  - Higher credit growth reduces the NPL ratio.
  - A one-standard deviation increase in the NPL ratio (an increase by 2.1 percentage point) has a negative effect on credit growth.
- Contribution to variance:
  - Forecast error variance decomposition: only 5 to 7 percent of the non-oil GDP growth variance can be explained by NPLs shocks (Figure 6).
- Selected impulse-response ranges (Monte-Carlo with 500 reps; Errors are 10% on each side):
  - Figure 7 (r, Credit, NPL, 2-lag VAR)
    - response of r to r shock: (p 10) r = -0.4486, (p 90) r = 1.3341
    - response of Credit to Credit shock: (p 10) Credit = -7.5113, (p 90) Credit = 13.1080
    - response of NPL to NPL shock: (p 10) NPL = -0.7179, (p 90) NPL = 2.5254
  - Figure 8 (r, Y, NPL, 2-lag VAR)
    - response of Y to NPL shock: (p 10) NPL = -1.2178, (p 90) NPL = 0.8879
    - response of NPL to Y shock: (p 10) Y = -1.0616, (p 90) Y = 0.4207
    - response of NPL to NPL shock: (p 10) NPL = 0.0000, (p 90) NPL = 2.1051
  - Figure 9 (Difference model, r, Y, NPL, 2-lag VAR)
    - response of Y to NPL shock: (p 10) NPL = -3.5982, (p 90) NPL = 0.7205
    - response of NPL to NPL shock: (p 10) NPL = -0.4666, (p 90) NPL = 2.4843
  - Figure 10 (r, Y, NPL, CPI, 2-lag VAR)
    - response of CPI to r shock: (p 10) r = -1.2781, (p 90) r = 0.3799
    - response of CPI to NPL shock: (p 10) NPL = -1.0460, (p 90) NPL = 0.3053
    - response of NPL to CPI shock: (p 10) CPI = -0.6544, (p 90) CPI = 1.3614
- Additional quantitative points:
  - The opening line reports effects of "1.5 and 2.2 percentage points after two and three years, and the effect is significant in the third period."
  - A 2 percentage point increase in the NPL ratio ⇒ a reduction of non-oil GDP growth by 0.8 percentage point one year after the shock.

### Robustness and caveats
- Results are tentative given small data set and period without systemic banking crises.
- Feedback effects likely nonlinear and could be much larger once NPLs cross critical thresholds.
- VAR in differences and adding CPI as a fourth variable produced broadly comparable feedback findings, though adding variables in small VARs reduces degrees of freedom and significance for several parameters.

### Policy recommendations and implications
- Strengthen macroprudential regulation given exchange rate peg constraints:
  - Reinforce capital buffers.
  - Enhance liquidity buffers.
  - Implement countercyclical provisioning norms.
- Supervisory focus: monitor credit growth, bank efficiency metrics, and external financing conditions (VIX) as key early indicators of NPL buildup.
- Use estimated elasticities (from bank-level models) as inputs into macroeconomic stress tests to assess banks’ resilience and potential feedback to the real economy.

*Source: IMF staff analysis based on Bankscope bank-level data and GCC macroeconomic series as presented in "Feedback Effect, 4 Variable Level Model" (section 10) of the supplied IMF working paper excerpt.*

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

### _wp10224 - References .............................................................................................................

### Tables
- 1.  Summary Statistics on Nonperforming Loans .................................................................... 4
- 2.  Macroeconomic and Firm-Specific Determinants of NPLs........................ .............. 10
- 3.  Panel Unit Root Tests (Levin-Lin-Chu) ........................................................................... 13
- 4.  Bankscope Coverage in the GCC ..................................................................................... 20

### Figures
- 1.   NPL Ratio and Economic Activity in the GCC .................................................................. 5
- 2.   Bank Heterogeneity and the Business Cycle ...................................................................... 6
- 3.   Logit Transformation of the NPL ratio ............................................................................... 9
- 4.   Dynamics of NPLs with Maintained Macroeconomic Shocks ......................................... 12
- 5.   Banking System Nonperforming Loans............................................................................ 13
- 6.   Variance Decomposition, Non-oil Real GDP ................................................................... 14
- 7.   Feedback Effect with Credit, 3 Variable Level Model ..................................................... 21
- 8.   Feedback Effect with GDP, 3 Variable Level Model ....................................................... 22
- 9.   Feedback Effect, 3 Variable Difference Model ................................................................ 23

*Source: _wp10224 - References .............................................................................................................*

### 10. Feedback Effect, 4 Variable Level Model ........................................................................ 24

### 10. Feedback Effect, 4 Variable Level Model

### Introduction
- Focus: relationship between macroeconomic variables and NPLs (credit risk) in GCC banks’ books using bank-level data from Bankscope.
- Sample/coverage: dynamic panel estimated over 1995–2008 on around 80 banks in the GCC region.
- Key methodological elements:
  - Dependent variable: logit transformation of the NPLs ratio (log(NPLs/(1-NPLs))).
  - Models estimated: OLS, fixed effects, Arellano-Bond difference GMM, System GMM (with collapsed instruments and forward orthogonalization).
  - Macroeconomic panel VAR estimated to investigate feedback effects of bank performance on credit supply and growth.
- Main empirical summary:
  - NPL ratios worsen as economic growth becomes lower and interest rates increase.
  - Larger banks and banks with lower expenses tend to have lower NPLs.
  - High past credit growth can generate higher NPLs in the future.
  - NPLs are very persistent; AR coefficient of logit(NPL) estimated between 0.6 and 0.9.
  - Cumulative effect of macroeconomic shocks over a three year horizon is large.
- Policy implication highlighted: in the context of exchange rate pegs, a stronger focus on macroprudential regulation—particularly through capital and liquidity buffers and countercyclical provisioning norms—would help mitigate macroeconomic risks on the banking system and the feedback effect of credit risks on the economy.

### Stylized facts on GCC banking system
- Historical context:
  - GCC experienced significant increases in banking system credit between 2003 and 2008; favorable pre-crisis macro environment associated with lower NPLs.
  - NPLs reached very high levels in the GCC before the boom years; NPL ratios in double-digits were not uncommon historically.
- 2009 country NPL increases (reported):
  - Bahrain 3.9 percent, Kuwait 9.7 percent, Oman 2.8 percent, Qatar 1.7 percent, Saudi Arabia 3.3 percent, U.A.E. 4.6 percent.
- Summary statistics (Bankscope, 1995–2008, unweighted NPL ratio in 2008 and percentiles):
  - NPL ratio in 2008 (unweighted): Bahrain 2.9, Kuwait 5.5, Oman 1.2, Qatar 1.7, Saudi Arabia 1.4, UAE 2.4
  - NPL ratio (1995–2008) percentiles (by country median unless otherwise noted):
    - 25th percentile: Bahrain 2.1, Kuwait 3.4, Oman 3.5, Qatar 1.3, Saudi Arabia 1.7, UAE 1.7
    - Median: Bahrain 5.9, Kuwait 6.7, Oman 6.5, Qatar 3.1, Saudi Arabia 3.1, UAE 4.3
    - 75th percentile: Bahrain 14.8, Kuwait 12.5, Oman 11.3, Qatar 11.5, Saudi Arabia 7.3, UAE 9.6
- Heterogeneity: in downturns, NPLs increase much faster for banks with higher initial NPL levels (nonlinear amplification across banks).

### Literature (determinants and feedbacks)
- Macroeconomic determinants:
  - Literature confirms macroeconomic conditions matter for credit risk (examples cited include Keeton and Morris (1987); asset-price links in Mueller (2000); Anderson and Sundaresan (2000); Collin-Dufresne and Goldstein (2001)).
  - Relevant macro variables: GDP growth, interest rates, exchange rate, unemployment, asset and house prices; paper focuses on non-oil real GDP growth, stock returns, interest rates, world trade growth, VIX index, and a 1997–1998 Asian crisis dummy.
- Bank-specific determinants:
  - Bank size, capital ratios, operating efficiency and past credit growth are relevant (literature examples: Salas and Saurina (2002); Bercoff, Giovanni and Grimard (2002)).
  - Moral hazard, agency problems, ownership structure, regulatory actions influence risk-taking; operational efficiency interacts with monitoring and loan quality.
- Feedback effects:
  - Concern that higher NPLs and impaired bank balance sheets constrain credit supply and slow recovery (cited findings: Keeton (1999); Lis et al. (2000); von Peter (2004); Peek and Rosengreen (2000); Dell’Ariccia et al. (2005); Ciccarelli et al. (2010)).
  - Four feedback channels summarized in literature: interbank contagion; correlation between credit and market risks; asset-price ↔ balance-sheet interactions; financial shock → real economy via tighter credit (Kida, 2008). This study focuses on the last channel.

### Determinants of credit risk — bank-level panel results
- Model setup:
  - Dependent variable: logit(NPL ratio).
  - Explanatory macro variables: non-oil real GDP growth, stock market returns, interest rates, world trade growth, VIX, 1997–1998 dummy.
  - Firm-level controls: capital adequacy ratio, measures of efficiency (expenses/asset, cost/income, ROE), ln(equity) (bank size), lagged net interest margin, lagged credit growth (deflated by CPI).
- Key estimated coefficients (selected results from Table 2 across specifications):
  - Autoregressive term ln(NPL/(1-NPL)) lagged: OLS 0.898***, FE 0.676***, Arellano-Bond 0.714***, System GMM (collapsed) 0.881***, System GMM (fwd. orth., collapsed) 0.865***.
  - ln(equity) lagged: OLS -0.0543**, FE -0.364***, Arellano-Bond -0.359***, System GMM -0.102, System GMM (fwd) -0.102.
  - (expenses/avg. assets) lagged: OLS 0.0483*, FE 0.0868*, Arellano-Bond 0.154**, System GMM 0.114**, System GMM (fwd) 0.106**.
  - loans growth lagged: OLS 0.104**, FE 0.0946*, Arellano-Bond 0.0993*, System GMM 0.137, System GMM (fwd) 0.145.
  - non-oil GDP growth lagged: OLS -1.948***, FE -0.974, Arellano-Bond -1.893**, System GMM -2.156*, System GMM (fwd) -2.090*.
  - interest rate lagged: OLS 0.0241*, FE 0.0535**, Arellano-Bond 0.0044, System GMM -0.0053, System GMM (fwd) 0.0001.
  - VIX (global risk aversion) contemporaneous/lagged: OLS 0.0131***, FE 0.0115***, Arellano-Bond 0.0119***, System GMM 0.0140***, System GMM (fwd) 0.0132***.
- Model diagnostics and sample:
  - Observations: 426 (columns where reported).
  - Number of banks in sample: 79 (varies by specification where noted).
  - No. of instruments in GMM: kept below 40; Hansen test p-values reported (for example 0.27, 0.48, 0.52); A-B AR(1) test p-values 0.00; A-B AR(2) test p-values 0.17–0.19.
- Interpretations and quantitative implications:
  - Autocorrelation/persistence: estimated AR coefficient between 0.6 (FE biased downward) and 0.9 (OLS biased upward) implies slow response to shocks but potential cumulation of NPLs over time.
  - Macroeconomic shocks: a temporary decrease by 3 percentage points in non-oil GDP growth would increase NPLs by 0.3 to 1.1 percentage points, depending on initial NPL levels (non-linearity of logit model implies marginal effects vary across the distribution).
  - A 300 basis points increase in interest rates would have a broadly similar effect on NPLs.
  - VIX is highly significant in all specifications, indicating external financing conditions/global risk aversion matter for GCC credit risk.
  - World trade growth and the Asian crisis dummy were not significant in these specifications.
  - Lagged credit growth is associated with higher future NPLs (evidence of balance-sheet expansion driving future credit risk).
  - Non-linearities: shocks have larger effects for banks with lower-quality loan portfolios; banks with higher expenses or faster expansion were tested for higher sensitivity but interaction terms were not significant—likely already captured by logit non-linearity.

### Feedback from the banking system to the real economy (Panel VAR results)
- Purpose: estimate the effect of increases in NPLs on credit growth and non-oil GDP growth using a panel VAR for GCC banking-system NPLs.
- Data notes:
  - Banking-system NPL data available since late 1990s for countries (examples: Bahrain since 1999; Kuwait since 1996; Oman since 1997; Qatar since 1999; Saudi Arabia since 1997; U.A.E. since 1996).
  - Small and unbalanced panel; unit-root tests (Levin-Lin-Chu) results not robust to sample period.
- Stationarity/unit-root testing (Levin-Lin-Chu test results, selected):
  - NPLs/Total Loans 1999-2008: 12.7*** (level), NPLs/Total Loans 2000-2008: -0.6 (level), difference -4.6***.
  - Non-oil real GDP growth 1998-2008: -3.3*** (level), 1999-2008: -0.7 (level), difference -1.4*.
  - Credit growth 1998-2008: 1.9 (level), difference -3.2***.
  - Interest rate 1998-2008: -6.2*** (level).
  - Significance levels: * 10%; ** 5%; *** 1%.
- VAR specification and identification:
  - Two panel VARs estimated: one with NPL ratio in levels, one with NPL ratio in first differences.
  - Lag length restricted to 2 lags (time dimension limited to 9 years).
  - Variables demeaned using Helmert procedure.
  - Choleski identification ordering: interest rate first, credit growth (or non-oil real GDP growth) second, NPL ratio last — reflecting pegged exchange rate regimes and assumption that NPLs do not affect contemporaneous interest or credit growth.
- Key empirical findings:
  - Panel VAR suggests a strong albeit short-lived feedback effect of higher NPLs on non-oil growth, reported as a semi-elasticity of around 0.4.
  - In the VAR with NPLs in levels (NPL ordered last) and including interest rate and change in credit growth:
    - Higher interest rates increase NPLs (effect not always significant).
    - Higher credit growth reduces the NPL ratio (row 3, columns 1 and 2 in reported results).
    - A one-standard deviation increase in the NPL ratio (an increase by 2.1 percentage point) has a negative effect on credit growth (panel VAR impulse-response results presented in the report).
  - Caution: results are tentative given small data set and period without systemic banking crises; feedback effects likely nonlinear and could be much larger once NPLs cross critical thresholds.
- Implication: banking-sector distress can reduce credit growth and weigh on non-oil GDP growth, suggesting the importance of policies that limit the build-up of NPLs and strengthen banks’ shock-absorption capacity.

### Policy recommendations and implications
- Strengthen macroprudential regulation given exchange rate peg constraints:
  - Reinforce capital buffers.
  - Enhance liquidity buffers.
  - Implement countercyclical provisioning norms.
- Supervisory focus: monitor credit growth, bank efficiency metrics, and external financing conditions (VIX) as key early indicators of NPL buildup.
- Use estimated elasticities (from bank-level models) as inputs into macroeconomic stress tests to assess banks’ resilience and potential feedback to the real economy.

*Source: IMF staff analysis based on Bankscope bank-level data and GCC macroeconomic series as presented in "Feedback Effect, 4 Variable Level Model" (section 10) of the supplied IMF working paper excerpt.*

### 1.5 and 2.2 percentage points after two and three years, and the effect is significant in the

### _wp10224 - 1.5 and 2.2 percentage points after two and three years, and the effect is significant in the

### Key empirical findings on NPL determinants
- Both macro-factors and bank-specific characteristics determine the level of nonperforming loans (NPLs) in the GCC banking sector.
- There is a significant inverse relationship between real (non-oil) GDP and nonperforming loans.
- Global financial market conditions have an effect on NPLs of banks.
- Among bank control factors, efficiency and past expansion of the balance sheet were found to be significant.

### VAR results and feedback from NPLs to real activity
- A panel VAR was estimated to assess feedback from rising NPLs to the real economy.
- A 2 percentage point increase in the NPL ratio was found to reduce non-oil GDP growth by 0.8 percentage point a year after the shock, and the feedback effect was significant (including at the 95 percent confidence interval).
- The model suggests a strong, though short-lived, feedback effect of default on non-oil growth, with a semi-elasticity of around 0.4.
- Default shocks do not occur often: the forecast error variance decomposition shows that only 5 to 7 percent of the non-oil GDP growth variance can be explained by NPLs shocks (Figure 6).
- Specific impulse-response findings (selected exact ranges shown in figures):
  - Figure 7 (r, Credit, NPL, 2-lag VAR; Monte-Carlo with 500 reps; Errors are 10% on each side)
    - response of r to r shock: (p 10) r = -0.4486, (p 90) r = 1.3341
    - response of Credit to Credit shock: (p 10) Credit = -7.5113, (p 90) Credit = 13.1080
    - response of NPL to NPL shock: (p 10) NPL = -0.7179, (p 90) NPL = 2.5254
  - Figure 8 (r, Y, NPL, 2-lag VAR; Monte-Carlo with 500 reps; Errors are 10% on each side)
    - response of Y to NPL shock: (p 10) NPL = -1.2178, (p 90) NPL = 0.8879
    - response of NPL to Y shock: (p 10) Y = -1.0616, (p 90) Y = 0.4207
    - response of NPL to NPL shock: (p 10) NPL = 0.0000, (p 90) NPL = 2.1051
  - Figure 9 (Difference model, r, Y, NPL, 2-lag VAR; Monte-Carlo with 500 reps; Errors are 10% on each side)
    - response of Y to NPL shock: (p 10) NPL = -3.5982, (p 90) NPL = 0.7205
    - response of NPL to NPL shock: (p 10) NPL = -0.4666, (p 90) NPL = 2.4843
  - Figure 10 (r, Y, NPL, CPI, 2-lag VAR; Monte-Carlo with 500 reps; Errors are 10% on each side)
    - response of CPI to r shock: (p 10) r = -1.2781, (p 90) r = 0.3799
    - response of CPI to NPL shock: (p 10) NPL = -1.0460, (p 90) NPL = 0.3053
    - response of NPL to CPI shock: (p 10) CPI = -0.6544, (p 90) CPI = 1.3614

### Robustness checks and additional identification tests
- A similar VAR using real GDP growth instead of growth in credit produced very comparable results: higher interest rates increase NPLs and higher non-oil growth reduces NPLs; both effects were found to be significant (see Figure 8).
- The model in differences (NPL ratio and growth rate in difference) again shows a significant feedback effect, although the effects of interest rate and GDP shocks on NPLs disappear (Figure 9).
- Adding consumer price index (CPI) as a fourth variable reduced significance for several variables (adding a fourth variable in a VAR with around 50 degrees of freedom is risky), but:
  - The feedback effect from NPLs to growth remained significant.
  - Inflation moved in the expected directions: inflation decreases after a tightening of monetary policy, but increases when activity is above trend — and both effects are significant (Figure 10).

### Magnitudes and timing highlighted
- The opening line reports effects of "1.5 and 2.2 percentage points after two and three years, and the effect is significant in the third period."
- Semi-elasticity of NPLs on non-oil growth: around 0.4.
- Contribution of NPL shocks to non-oil GDP growth variance: 5 to 7 percent.
- A 2 percentage point increase in the NPL ratio ⇒ a reduction of non-oil GDP growth by 0.8 percentage point one year after the shock.

### Policy implications and recommendations
- Policymakers in the GCC should monitor carefully the evolution of default in bank loan books because feedback effects, though normally short-lived, could become nonlinear and lead to substantially larger costs if NPLs cross certain thresholds.
- In the context of exchange rate pegs, a stronger focus on macroprudential regulation is warranted, particularly through:
  - capital and liquidity buffers,
  - countercyclical provisioning,
  to mitigate the impact of macroeconomic risks on the banking system and the feedback effects of credit risks on the economy.

*Source: IMF Working Paper content from the provided PDF excerpt.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10224.pdf_