## wp1766 - References

## Source details

**Canonical URL:** [wp1766 - References](https://www.imf.org/-/media/files/publications/wp/2017/wp1766.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1766.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1766.pdf.json)

---

### Introduction and context
- The ratio of non-performing loans (NPLs) to total loans in Italy reached very high levels post global financial crisis and "Total NPLs appear to have broadly stabilized at aboute356 billion at end-June 2016 (about 18 percent of total loans; 20 percent of GDP; and one-third of the Euro Area total)".
- Regional concentration: NPLs "as a ratio of total assets, they were mostly concentrated in the south of the country".
- Net of provisions: "current NPLs amount to about 191 billion (10.4 percent of total loans); out of these, 88 billion (4.8 percent of loans) are represented by bad loans ... the remaining 103 billion relate to situations in which repayments may still resume."
- Channels through which high NPLs adversely affect activity:
  - Banks focus on internal consolidation and asset quality rather than new credit.
  - Higher loan loss provisions reduce resources for lending.
  - Misallocation of resources (zombie lending, evergreening).

### Research question, data, and sample
- Core question: "Can Italy grow out of its NPL overhang?"
- Approach: investigate a possible non-monotonic relationship between real GDP growth and the NPL ratio in Italy, accounting for feedback from NPLs to growth.
- Data: regional NPL ratios from Bank of Italy supervisory returns and regional real GDP from Istat.
- Sample: a panel of 17 Italian regions over the period 1997–2014.
- Advantages of within-country regional analysis: avoids cross-country heterogeneity and varying NPL definitions; allows for region-level heterogeneity, dynamics, and feedback effects.

### Key empirical findings
- Statistically significant growth-threshold:
  - "We find a statistically significant growth-threshold effect on the NPL ratio in Italy at about1:2 percent" once accounting for heterogeneities, simultaneity, and dynamics.
  - Table 1 reports threshold estimates b = 1.2% across ARDL and DL specifications and lag choices.
  - SupT and AveT test statistics (selected entries) reported in the source.
- Long-run negative association between growth and NPL ratios:
  - "Quantitatively, a one percentage points faster growth than the baseline in Italy, if persistent, would reduce the NPL ratio by about6:5to9:5percent per year (i.e. halving the NPL ratio in3 6years)."
- Robustness:
  - Sup and Ave tests are statistically significant "in all cases, irrespective of the choice of the lag order and the estimation procedure (ARDL or DL)."
- Mechanisms for non-linearity:
  - Faster growth raises collateral values and narrows market-book value gaps for NPLs, encouraging disposals and write-offs.
  - Sustained growth can restore borrowers’ repayment capacity and reduce new defaults.
  - Higher profitability during sustained growth allows banks to use retained earnings to build capital and write off NPLs.

### Econometric model and methods
- Baseline dynamic panel-threshold ARDL specification for changes in log NPL ratio with threshold indicator I(y_it > ) and lag structure; augmented by a real GDP growth equation allowing feedback from lagged NPL growth to growth.
- Identification strategy: model correlation between innovations and derive reduced-form panel threshold-ARDL where threshold variable is excluded from growth equation.
- Testing procedure: panel SupF/SupT and AveF/AveT statistics following Chudik et al. (2017) with grid search for threshold parameter ; square-root transformation used for single threshold variable (r = 1).
- Lag choices: considered p up to p_max = 2; ARDL and DL counterparts estimated (equations (8) and (9)).
- Estimation: Least squares Mean Group (MG) estimates of long-run effects reported; MG estimates are consistent under fairly general conditions so long as the errors are cross-sectionally independent.

### Robustness to nominal dynamics
- Re-estimated model using nominal GDP growth.
- Obtained a "nominal GDP growth" threshold estimate of about 3 percent (Table 3).
- Noted that the GDP deflator in Italy grew on average by about 2 percent over 1997-2014.
- Given projected inflation in Italy remains significantly below the European Central Bank’s target of 2 percent, decisive reduction in NPL overhang relies on improving underlying real growth dynamics (i.e. real GDP growth above 1.2 percent).

### Policy implications and recommendations
- Main conclusion: there is a growth-threshold effect of about 1.2 percent in the relationship between real GDP growth and NPL ratios for Italy.
  - Persistent real GDP growth above 1.2 percent for a number of years is needed to reduce NPL ratios significantly over the medium term.
- Given Italy’s moderate growth outlook ("real GDP growth ... to remain close to 1 percent over the next few years"), relying solely on cyclical growth may not suffice to eliminate the NPL overhang.
- Active NPL resolution measures needed:
  - More intensive use of out-of-court debt restructuring mechanisms.
  - Strengthened supervision.
  - Systematic assessment of asset quality for banks not already subject to the ECB comprehensive assessment, with follow-up actions in line with regulatory requirements.
  - Effective use of the framework for the prompt resolution of banks, addressing concerns related to the bail-in of retail investors appropriately.
- Structural reforms and fiscal adjustments to boost growth:
  - Fully implement and scale up recent reform efforts.
  - Lower Italy’s high levels of public debt.
  - Ensure a pro-growth mix of spending and tax measures.
- Insolvency reforms: "the insolvency reforms, once fully implemented, are expected to yield benefits only gradually over time," so complementary measures are required to address the existing stock of NPLs.

### Conclusions
- Evidence supports a non-monotonic relationship between real GDP growth and the NPL ratio in Italy, with a threshold near1:2 percent growth needed to trigger significant NPL reductions.
- Persistent faster growth would materially reduce NPL ratios ("about6:5to9:5percent per year for a 1 percentage point faster growth"), but given Italy’s modest growth projections, active bank-sector and policy measures remain necessary to resolve the NPL overhang within a reasonable time frame.

### Key statistics and exact figures reported
- Growth threshold: 1.2 percent (real GDP growth).
- Sample period: 1997–2014.
- Nominal GDP growth threshold estimate: about 3 percent.
- GDP deflator average growth in Italy over 1997-2014: about 2 percent.
- Estimated long-run coefficients of real GDP growth on changes in NPL ratios (b): range from  6:5to 9:5 (statistically significant at the 1 percent level).
- Effect interpretation: a one percentage point faster growth than the baseline reduces NPL ratio by about6:5to9:5percent per year (halving NPL stock in3 6years).
- Reported NPL stock context: "Total NPLs appear to have broadly stabilized at aboute356 billion at end-June 2016 (about 18 percent of total loans; 20 percent of GDP; and one-third of the Euro Area total)".
- Net of provisions: "current NPLs amount to about 191 billion (10.4 percent of total loans); out of these, 88 billion (4.8 percent of loans) are represented by bad loans ... the remaining 103 billion relate to situations in which repayments may still resume."

*Source: wp1766 - References (PDF chapter/section).*

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

### wp1766 - References

### Introduction and context
- The ratio of non-performing loans (NPLs) to total loans in Italy reached very high levels post global financial crisis and "Total NPLs appear to have broadly stabilized at aboute356 billion at end-June 2016 (about 18 percent of total loans; 20 percent of GDP; and one-third of the Euro Area total)".
- Regional concentration: NPLs "as a ratio of total assets, they were mostly concentrated in the south of the country".
- Net of provisions: "current NPLs amount to about 191 billion (10.4 percent of total loans); out of these, 88 billion (4.8 percent of loans) are represented by bad loans ... the remaining 103 billion relate to situations in which repayments may still resume."
- Channels through which high NPLs adversely affect activity:
  - Banks focus on internal consolidation and asset quality rather than new credit.
  - Higher loan loss provisions reduce resources for lending.
  - Misallocation of resources (zombie lending, evergreening).

### Research question, data, and sample
- Core question: "Can Italy grow out of its NPL overhang?"
- Approach: investigate a possible non-monotonic relationship between real GDP growth and the NPL ratio in Italy, accounting for feedback from NPLs to growth.
- Data: regional NPL ratios from Bank of Italy supervisory returns and regional real GDP from Istat.
- Sample: a panel of 17 Italian regions over the period 1997–2014.
- Advantages of within-country regional analysis: avoids cross-country heterogeneity and varying NPL definitions; allows for region-level heterogeneity, dynamics, and feedback effects.

### Key empirical findings
- Statistically significant growth-threshold:
  - "We find a statistically significant growth-threshold effect on the NPL ratio in Italy at about1:2 percent" once accounting for heterogeneities, simultaneity, and dynamics.
  - Table 1 (Tests of real GDP growth-threshold effects on changes in NPL ratios) reports threshold estimates b = 1.2% across ARDL and DL specifications and lag choices.
  - SupT and AveT test statistics (selected entries):
    - SupT: 4.31 z; 4.54 z; 1.89 3.11  3.11  (as reported in the table).
    - AveT: 3.00 z; 3.13 z; 1.17 ; 1.88 z; 1.98 z (as reported).
- Long-run negative association between growth and NPL ratios:
  - "Quantitatively, a one percentage points faster growth than the baseline in Italy, if persistent, would reduce the NPL ratio by about6:5to9:5percent per year (i.e. halving the NPL ratio in3 6years)."
- Robustness:
  - Sup and Ave tests are statistically significant "in all cases, irrespective of the choice of the lag order and the estimation procedure (ARDL or DL)."
- Mechanisms for non-linearity:
  - Faster growth raises collateral values and narrows market-book value gaps for NPLs, encouraging disposals and write-offs.
  - Sustained growth can restore borrowers’ repayment capacity and reduce new defaults.
  - Higher profitability during sustained growth allows banks to use retained earnings to build capital and write off NPLs.

### Econometric model and methods
- Baseline dynamic panel-threshold ARDL specification for changes in log NPL ratio (d_it) with threshold indicator I(y_it > ) and lag structure; augmented by a real GDP growth equation allowing feedback from lagged NPL growth to growth.
- Identification strategy: model correlation between innovations and derive reduced-form panel threshold-ARDL where threshold variable is excluded from growth equation.
- Testing procedure: panel SupF/SupT and AveF/AveT statistics following Chudik et al. (2017) with grid search for threshold parameter ; square-root transformation used for single threshold variable (r = 1).
- Lag choices: considered p up to p_max = 2; ARDL and DL counterparts estimated (equations (8) and (9)).

### Policy implications and recommendations
- Given Italy’s moderate growth outlook (with various analysts projecting "real GDP growth ... to remain close to 1 percent over the next few years"), relying solely on cyclical growth may not suffice to eliminate the NPL overhang.
- Active NPL resolution measures are needed to put NPL ratios on a firm downward trajectory over the medium term. Measures discussed include:
  - More intensive use of out-of-court debt restructuring mechanisms.
  - Strengthened supervision.
  - Systematic assessment of asset quality for banks not already subject to the ECB comprehensive assessment, with follow-up actions in line with regulatory requirements.
  - Effective use of the framework for the prompt resolution of banks, addressing concerns related to the bail-in of retail investors appropriately.
- Structural reforms and fiscal adjustments to boost growth are critical:
  - Fully implement and scale up recent reform efforts.
  - Lower Italy’s high levels of public debt.
  - Ensure a pro-growth mix of spending and tax measures.
- Expectation on insolvency reforms: "the insolvency reforms, once fully implemented, are expected to yield benefits only gradually over time," so complementary measures are required to address the existing stock of NPLs.

### Conclusions
- Evidence supports a non-monotonic relationship between real GDP growth and the NPL ratio in Italy, with a threshold near1:2 percent growth needed to trigger significant NPL reductions.
- Persistent faster growth would materially reduce NPL ratios (about6:5to9:5percent per year for a 1 percentage point faster growth), but given Italy’s modest growth projections, active bank-sector and policy measures remain necessary to resolve the NPL overhang within a reasonable time frame.

*Source: wp1766 - References (PDF chapter/section).*

### 1.2 percent over the medium term (see Figure 4). Slow growth has prevailed despite important

### wp1766 - 1.2 percent over the medium term (see Figure 4). Slow growth has prevailed despite important

### Context and need for reforms
- Slow growth has prevailed despite important reform efforts over the past two decades, owing in part to weaknesses in implementation.
- The potential real GDP growth in Italy exceeded 1.2 percent only before the millennium, and both long-term average real GDP and TFP growth rates are well below those of similar economies.
- Further efforts are needed in three important areas to raise potential growth and help reduce NPL ratios faster:
  - product and service markets;
  - public administration;
  - wage bargaining reform to align wages with productivity at the firm level and across regions.
- Active resolution of NPLs is needed as outlined in Section I.
- Ensuring a pro-growth mix of spending and tax measures would also help.

### Estimates of long-run effects
- Methodology:
  - Used ARDL and DL specifications (equations (8) and (9)) following Pesaran and Smith (1995), Pesaran (1997), and Pesaran and Shin (1999).
  - Panel ARDL approach valid regardless of regressors being exogenous or endogenous and whether variables are I(0) or I(1).
  - Both ARDL and DL allow for cross-region heterogeneity and varying short-run effects across regions.
- Estimation details:
  - Least squares Mean Group (MG) estimates of the long-run effects of real GDP growth, y_it, on changes in NPL ratios are reported in Table 2.
  - MG estimates are consistent under fairly general conditions so long as the errors are cross-sectionally independent.
- Main empirical finding:
  - The coefficients of real GDP growth, b, are negative and statistically significant at the 1 percent level.
  - Coefficient values range from  6:5to 9:5 across ARDL and DL specifications and lag orders.
  - Interpretation: a one percentage point faster growth than the baseline, if it persists, would reduce the ratio of NPLs by about6:5to9:5percent per year (i.e., halving the NPL stock in3 6years).

### Robustness to inflation dynamics
- Re-estimated model using nominal GDP growth.
- Obtained a "nominal GDP growth" threshold estimate of about 3 percent (Table 3).
- Noted that the GDP deflator in Italy grew on average by about 2 percent over 1997-2014.
- Given projected inflation in Italy remains significantly below the European Central Bank’s target of 2 percent, decisive reduction in NPL overhang relies on improving underlying real growth dynamics (i.e. real GDP growth above 1.2 percent).

### Concluding remarks and policy recommendations
- Main conclusion: there is a growth-threshold effect of about 1.2 percent in the relationship between real GDP growth and NPL ratios for Italy.
  - Persistent real GDP growth above 1.2 percent for a number of years is needed to reduce NPL ratios significantly over the medium term.
- Requirements to achieve growth above 1.2 percent:
  - Tackling long-standing structural rigidities.
  - Lowering Italy’s high levels of public debt.
  - Ensuring a pro-growth mix of spending and tax measures.
- Financial sector measures needed to clean up bank balance sheets:
  - More intensive use of out-of-court debt restructuring mechanisms.
  - Strengthened supervision to facilitate decisive progress in reducing NPLs.
  - Additional measures beyond those already introduced by the authorities could help bring down NPL ratios faster.

### Key statistics and exact figures reported
- Growth threshold: 1.2 percent (real GDP growth).
- Sample period: 1997–2014.
- Nominal GDP growth threshold estimate: about 3 percent.
- GDP deflator average growth in Italy over 1997-2014: about 2 percent.
- Estimated long-run coefficients of real GDP growth on changes in NPL ratios (b): range from  6:5to 9:5 (statistically significant at the 1 percent level).
- Effect interpretation: a one percentage point faster growth than the baseline reduces NPL ratio by about6:5to9:5percent per year (halving NPL stock in3 6years).

*Source: IMF working paper excerpt (wp1766).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1766.pdf_