## 1itaea2020008

## Source details

**Canonical URL:** [1itaea2020008](https://www.imf.org/-/media/files/publications/cr/2020/english/1itaea2020008.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2020/english/1itaea2020008.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2020/english/1itaea2020008.pdf.json)

---

### Context and scope
- FSAP conducted a comprehensive set of stress tests and interconnectedness analyses to assess resilience and vulnerabilities of the Italian financial banking system.
- Coverage:
  - Solvency stress tests: 9 significant institutions (SIs), representing 68 percent of total banking assets.
  - Sensitivity-based analysis: both SIs and 62 less significant institutions (LSIs).
  - Liquidity stress tests: LCR, NSFR, and cashflow-based analyses for 11 SIs and 61 LSIs.
  - Corporate sector stress tests and profitability analysis supplemented the banking tests.
- Note: cooperative banking sector not stress tested due to consolidation into two groups and a joint protection scheme in early January 2019.

### Key vulnerabilities and structural factors
- Legacy issues and improvements:
  - NPL ratio fell from 16.5 percent in 2015 to about 8.7 percent in December 2018; close to €130 billion of private NPL sales contributed to the reduction.
  - Shortfalls of loan loss provisions estimated at about €5 billion based on internal workouts (mostly related to loans identified as unlikely to pay).
  - Additional €7½ billion needed for banks to halve their NPLs through market sales.
  - Provisioning coverage increased by 2 percentage points in 2018 to 53 percent.
- Funding support and concentration risks:
  - Italian banks’ use of the ECB’s TLTRO is close to € 250 bn, representing on average about 10 percent of banks’ total liabilities.
  - Concentration of liquid assets in Italian government securities increases liquidity susceptibility to adverse market valuations.
- Profitability and governance:
  - Sample SIs fully-loaded CET1 ratio of 11.9 percent in December 2018 (0.3 percentage point increase over two years), 2.6 percentage points below EU peers.
  - Bank ROA: 0.6 percent in 2017 and 0.5 percent in 2018.
  - High operating costs in segments and corporate governance weaknesses weigh on profitability.
  - IFRS 9 and MREL implementation will further affect profitability and capital dynamics.
- Sovereign and corporate sector linkages:
  - Banks’ exposures to domestic sovereign over 11 percent of total assets (Domestic Government Bonds share as of November 2018: Italy 10.5).
  - BdI estimate: upward shift of 100 bps in the government yield curve would reduce banks’ CET1 ratio by 40 bps.
  - Corporate sector vulnerabilities amplify banking sector credit risk exposure.

### Solvency stress test results — Baseline and adverse scenarios
- Baseline scenario (aggregate sample):
  - Aggregate CET1 ratio declines by 56 basis points from 11.9 percent to 11.4 percent with “no NPL disposal”.
  - Including NPL disposal targets, aggregate CET1 declines by 102 bps to 10.9 percent and resulting shortfall in capital of about 0.05 percent of GDP (alternate baseline).
  - Under baseline with no NPL disposal: one bank falls below the T1 threshold and another below the CAR threshold; two banks see leverage ratio fall below 3 percent.
  - Credit risk is main contributor to baseline capital decline.
- Adverse scenario (severe but plausible):
  - Average fully loaded CET1 capital ratio declines from 11.9 percent in 2018 to 8.2 percent in 2021.
  - Three banks would see capital ratios drop below capital minimum requirements; these three banks comprise about 8.5 percent of the sample’s assets.
  - Resulting capital shortfall against capital thresholds is about 0.2 percent of GDP.
  - Capital needed to bring the CET1 ratio of the 9 SIs back to end-2018 level of 12 percent is about 2.2 percent of GDP.
  - Contributors to CET1 decline across sample banks:
    - Credit risk: about 5.3 percentage points of the decline.
    - Valuation losses from increased sovereign yields: 1.2 percentage points.
    - Net interest income (NII) losses related to higher funding costs: 0.9 percentage points.
    - Expansion of RWAs: 1.2 percentage points (from heightened credit risk, FX valuation effects, and projected lending).
    - Pre-provision revenue increases aggregate CET1 by 4.7 percentage points relative to the starting point.

### Sensitivity analysis and single-factor shocks
- LSIs vulnerabilities:
  - An increase in Italian sovereign yields by 230 bps would cause the capital of almost a quarter of the sample of LSIs by assets (10 banks) to fall below the 7 percent CET1 threshold.
  - Under an NPL shock, 14 LSIs (35 percent of the sample’s assets) would see their CET1 ratio fall below 7 percent.
- Debt portfolio sensitivity (LSIs, Scenario 1 — 2011 rapid rise in yields):
  - Capital losses amount to 300 bps for LSIs; almost a quarter of group by assets fall below 7 percent CET1.
  - Total capital shortfall of 0.02 percent of GDP; capital needed to restore starting CET1 (15.2 percent) about 2.1 percent of GDP.
- Loan-loss sensitivity (LSIs, most severe NPL-flow scenario):
  - Capital losses amounted to 365 bps for LSIs; 14 banks (35 percent by assets) fall below 7 percent CET1.
  - Capital needs to bring banks’ CET1 back to 7 percent: about 0.1 percent of GDP; to starting CET1 (15.2 percent): 0.26 percent of GDP.
- Concentration risk (SIs):
  - Simultaneous default of five largest borrowers reduces aggregate CET1 by 3.4 percentage points (from 11.9 percent to 8.5 percent) using 100 percent provisioning; no banks fall below 4.5 percent CET1 in that test.

### Liquidity stress tests — LCR, NSFR, and cashflow analyses
- Aggregate positions:
  - System-wide LCR: SIs 156%; LSIs 229%.
  - Under retail stress scenario: SIs stressed LCR 92%; LSIs stressed LCR 157%.
  - Cashflow-based tests indicate generally comfortable short-term liquidity but rising vulnerabilities over a three-month horizon.
- Key vulnerabilities:
  - Reliance on TLTRO close to € 250 bn (about 10 percent of total liabilities) substantially supports aggregate liquidity but creates replacement cost risk.
  - Concentration of CBC in domestic government securities: LSIs CBC 82 percent government items; SIs 67 percent.
  - Asset encumbrance (AE): average AE SIs 30 percent; average AE LSIs 22 percent; some LSIs AE above 40 percent.
- Cashflow-based stress outcomes:
  - One-month horizon:
    - SIs: CBC as share of assets can decline from 15 percent to 5 percent across scenarios; potential liquidity needs could reach 12 billion euros in most severe market scenario.
    - LSIs: CBC shortfalls range from €20 million (medium stress) to almost €1.1 billion (most severe); CBC as share of assets can decline from 21 percent to 11 percent across calibrations.
  - Three-month horizon:
    - SIs: under most severe market scenario average impact can amount to -18 percent of total assets; potential liquidity needs up to 100 billion euros.
    - LSIs: 32 out of 55 LSIs unable to maintain positive CBC in most severe market stress; potential liquidity needs could reach almost 11 billion euros.
  - Five-day horizon: all SIs and LSIs maintain positive CBC even in most severe scenario (SIs CBC declines to 7 percent of assets; LSIs remains about 17 percent).
- LCR bank counts under retail scenario:
  - SIs: Number with LCR<100% — 9 banks; share of group assets 91%; liquidity shortfall €16,768 million.
  - LSIs: Number with LCR<100% — 10 banks; share of group assets 26%; liquidity shortfall €1,341 million.

### Corporate sector stress test — design and results
- Dataset: CERVED firm-level database 2002–17 covering more than 10 million observations for 1.5 million firms.
- Stress-test design:
  - 3-year horizon; baseline keeps end-2017 outstanding firm debt constant; shocks calibrated to 3-year cumulative nominal GDP deviation between adverse and baseline.
  - Shocks: profit shock, interest rate shock, and combined profit+interest shock; medium scenario assumes half the value of severe shocks.
- Adverse scenario impacts (combined profit and interest shock):
  - Share of vulnerable firms (ICR<1) rises to 40 percent from 26 percent pre-shock.
  - Debt-at-risk rises to 38 percent from 23 percent pre-shock.
  - Results are consistent with banking sector stress tests where majority of losses emanate from corporate credit risk.
- Structural findings:
  - Corporate debt-at-risk concentrated in micro and large firms; large firms hold 51 percent of debt and 40 percent of debt-at-risk.
  - Sector concentration: manufacturing, construction and real estate, trade, and professional services account for about 70 percent of debt-at-risk.
  - Geographic concentration: north-west holds about 36 percent of corporate debt and almost half of debt-at-risk.

### Interconnectedness and cross-border linkages
- Domestic interbank contagion:
  - Interbank exposures among 72 banks amount to 1.05 percent of banking system assets and 1.6 percent of GDP.
  - Balance-sheet based network analysis indicates very limited contagion risk owing to small interbank exposures; simulations robust to parameter changes.
- Cross-sector contagion:
  - Flow-of-funds analysis highlights significant cross-sectoral contagion risk due to growing cross-sectoral exposures and heavy common exposure to government bonds by banks, insurers, and foreign investors.
  - Households indirectly exposed to sovereign risk via financial intermediaries.
- Cross-border linkages:
  - Total consolidated foreign claims of Italian banks were US$0.65 trillion in 2018:Q2; foreign banks’ claims on Italy about US$0.66 trillion.
  - Outward spillovers from Italy largest to Luxembourg, Turkey and Austria (in banking exposures); inward spillovers to Italy largest from France, Germany, the United States, the United Kingdom, and Spain.
  - Market-based connectedness (Diebold and Yilmaz): Italy is an important net contributor to global banking return connectedness; Italian sovereign CDS is the second largest net contributor to sovereign CDS markets.
  - Since May 2018, Italian markets and sovereign CDS have shifted from net shock transmitters to more net shock receivers.

### Key policy implications and recommendations (selected, with time frames where provided)
- Enhance banks’ capital levels to ensure all banks maintain adequate capital ratios under stress scenarios. (ECB, BdI) — Time frame: NT.
- Incentivize banks to diversify bond holdings included in their counterbalancing capacity to limit market/liquidity risks. (ECB, BdI) — Time frame: NT.
- Supervisory actions:
  - Conduct thorough supervisory review of banks’ business models and governance, deploying supervisory toolkit to achieve improvements, consolidation, or orderly winddowns as needed.
- Liquidity and funding:
  - Reduce concentration of liquid assets in own-sovereign securities; diversify bond exposures to non-Italy sovereign bonds and vary maturities and asset types.
  - Reduce reliance on TLTRO (close to € 250 bn and about 10 percent of banks’ total liabilities) and on less-stable wholesale funding, particularly for SIs.
  - Address high asset encumbrance, especially LSIs with AE above 40 percent, to preserve access to unsecured wholesale and central bank funding.
- Corporate sector resilience:
  - Monitor corporate sector vulnerabilities (debt-at-risk and ICR): policy to support structural improvements in profitability and investment would reduce banking credit risk over time.
- Communications and macroprudential stance:
  - Recognize stress-test limitations and heterogeneity with other exercises (EBA, ECB); use results as macroprudential tools not as macroeconomic forecasts.

_Italic: Source: IMF staff calculations and analysis as presented in the provided content._

### EXECUTIVE SUMMARY __________________________________________________________________________ 7

### EXECUTIVE SUMMARY

### Context and scope
- FSAP conducted a comprehensive set of stress tests and interconnectedness analyses to assess resilience and vulnerabilities of the Italian financial banking system.
- Coverage:
  - Solvency stress tests: 9 significant institutions (SIs), representing 68 percent of total banking assets.
  - Sensitivity-based analysis: both SIs and 62 less significant institutions (LSIs).
  - Liquidity stress tests: LCR, NSFR, and cashflow-based analyses for 11 SIs and 61 LSIs.
  - Corporate sector stress tests and profitability analysis supplemented the banking tests.
- Note: The FSAP did not stress test the cooperative banking sector since it was undergoing consolidation into two groups and a joint protection scheme in early January 2019.

### Key vulnerabilities and structural factors
- Legacy issues and improvements:
  - Significant progress tackling legacy nonperforming loans (NPLs) and improving solvency ratios.
  - Shortfalls of loan loss provisions estimated at about €5 billion based on internal workouts (mostly related to loans identified as unlikely to pay).
  - An additional €7½ billion will be needed for banks to halve their NPLs through market sales.
- Funding support and concentration risks:
  - Italian banks are the largest utilizers of the ECB’s TLTRO: close to € 250 bn and representing on average about 10 percent of banks’ total liabilities.
  - Concentration of liquid assets in Italian government securities increases liquidity susceptibility to adverse market valuations of these securities.
- Profitability and governance:
  - Many banks still suffer from relatively low capital levels, low profitability, and asset quality.
  - High operating costs in segments of the system and corporate governance weaknesses weigh on profitability.
  - Implementation of IFRS 9 and MREL will further affect profitability and capital dynamics.
- Sovereign and corporate sector linkages:
  - Banks’ exposure to the sovereign and leverage in the corporate sector increase vulnerability to downside shocks.
  - Fiscal vulnerabilities raise the risk of substantial economic contraction and rising credit spreads with strong negative repercussions for the financial sector.

### Solvency stress test results — Baseline and adverse scenarios
- Baseline scenario:
  - Aggregate CET1 ratio of sample banks declines by 56 basis points from 11.9 percent to 11.4 percent.
  - With no additional NPL disposal assumed, one bank falls below the T1 threshold and another bank falls below the capital adequacy ratio (CAR) threshold.
  - Including NPL disposal targets, the first bank would fall short of the three capital thresholds.
  - Aggregate CET1 ratio declines by 102 bps to 10.9 percent with a resulting shortfall in capital of about 0.05 percent of gross domestic product (GDP) (when including NPL disposal targets as alternate baseline).
  - Credit risk is the main contributor to the decline in capital ratios under the baseline.
- Adverse scenario (severe but plausible):
  - Average fully loaded CET1 capital ratio declines from 11.9 percent in 2018 to 8.2 percent in 2021.
  - Three banks would see their capital ratios drop below capital minimum requirements.
  - Resulting capital shortfall against capital thresholds is about 0.2 percent of GDP.
  - Capital needs to bring the CET1 ratio of the 9 SIs included in the stress scenario back to the end-2018 level of 12 percent is about 2.2 percent of the GDP.
  - Contributors to CET1 decline across sample banks:
    - Credit risk: about 5.3 percentage points of the decline.
    - Valuation losses from increased sovereign yields: 1.2 percentage points decline in CET1.
    - Net interest income (NII) losses related to higher funding costs: 0.9 percentage points.
    - Expansion of RWAs (from heightened credit risk, FX valuation effects, and projected lending) lowering CET1 by 1.2 percentage points.
    - Pre-provision revenue (mainly aggregate NII, non-interest income, and non-interest expenses) increases aggregate CET1 by 4.7 percentage points relative to the starting point.

### Sensitivity analysis and single-factor shocks
- LSIs vulnerabilities:
  - An increase in Italian sovereign yields by 230 bps would cause the capital of almost a quarter of the sample of LSIs by assets (10 banks) to fall below the 7 percent CET1 ratio threshold.
  - Under an NPL shock, 14 LSIs (35 percent of the sample’s assets) would see their CET1 ratio fall below 7 percent.
- SIs concentration risk:
  - SIs are resilient to concentration risk; they can withstand the simultaneous default of the five largest borrowers of their non-financial corporate exposures.

### Liquidity stress tests
- Three approaches: NSFR-based, LCR-based, and cashflow-based liquidity stress tests.
- Results indicate relatively comfortable positions overall.
- Key vulnerabilities:
  - Concentration of liquid assets in Italian government securities makes liquidity susceptible to adverse market valuations.
  - Aggregate liquidity performance is substantially supported by reliance on TLTRO (close to € 250 bn; about 10 percent of total liabilities).

### Corporate sector stress test
- Non-financial corporate sector remains sensitive to macroeconomic shocks.
- In the adverse macroeconomic scenario, combined profit and interest rate shocks would move the median interest coverage ratio and the share of firms-at-risk close to the levels of the 2008–09 and 2012 crises.
- Improvements in corporates’ debt servicing capacity have been driven largely by historically low interest rates; structural improvements in profitability have been insufficient.
- These corporate vulnerabilities align with banking sector stress test results where majority of losses emanate from corporate credit risk.

### Interconnectedness and cross-border linkages
- Domestic interbank contagion:
  - Network analysis suggests very limited contagion risk within the banking system owing to small interbank exposures.
- Cross-sector contagion risk:
  - Flow-of-funds analysis indicates significant cross-sectoral contagion risk because of:
    - Growing cross-sectoral exposures.
    - Significant exposure of banks, nonbanks (particularly insurance firms), and foreign investors to the government bond market.
    - Indirect links of households to sovereign risk intermediated through the financial system.
    - Direct links of banks, households, and nonresidents to corporate debt.
- Cross-border linkages:
  - Italy’s banking system is highly integrated cross-border; shocks originating externally are increasingly relevant.
  - While historically Italian markets and sovereign CDS were net shock transmitters, their role as net shock originators declined since May 2018 and they became more sensitive to external shocks due to foreign investor sales of Italian securities and domestic residents building net foreign asset positions.

### Recommendations from stress testing and risk assessment (selected)
- Further enhance banks’ capital levels, as appropriate, to ensure all banks maintain adequate capital ratios under stress scenarios. (ECB, BdI) — Time frame: NT
- Incentivize banks to diversify bond holdings included in their counterbalancing capacity to limit market/liquidity risks. (ECB, BdI) — Time frame: NT

*Prepared by Maral Shamloo, Irman Pardede, Purva Khera, Nour Tawk, Xiaodan Ding and Mindaugas Leika, from the IMF Monetary and Capital Markets Department, and Alvar Kangur from the IMF European Department for the 2020 Italy FSAP.*

### INTRODUCTION

### INTRODUCTION

### Financial system structure and composition
- Banks continue to dominate the Italian financial system despite significant growth of insurance firms and investment funds.
- As of June 2019, the number of banks in the financial system was about 170.
- In January 2019, about 240 of the 280 mutual banks were merged into two new banking groups, which have been classified as SIs; the remaining mutual banks will enter into an institutional protection scheme (IPS).
- The insurance sector: fourth largest in Europe and eighth largest in the world by premium income; number of insurers fell from 162 in 2007 to 100 as of June 2018.
- Share of assets of investment funds and other financial intermediaries in the financial system grew from 15 percent in 2011 to 18 percent.

### Non-performing loans (NPLs) and provisioning
- NPLs ratio of the banking sector fell from 16.5 percent in 2015 to about 8.7 percent in December 2018.
- Reduction achieved mainly through close to €130 billion of private NPL sales.
- Italy’s NPL ratio remains above the 3.2 percent average of the main EU banks.
- SIs are planning to further reduce NPLs to 7 percent by end-2020.
- New NPL formation has fallen to pre-crisis levels.
- Provisioning coverage increased by 2 percentage points in 2018 to 53 percent, placing Italy at 8 percentage points above the average of the main EU banks.

### Capitalization, market valuation, and profitability
- In December 2018, the sample of SIs reported a fully-loaded CET1 ratio of 11.9 percent (0.3 percentage point increase in the last two years), which is 2.6 percentage points below the average of their EU peers.
- Market-adjusted measures indicate aggregate price-to-book ratios less than one; using market valuations would reduce Italian banking sector capital ratios by about 45 percent.
- Bank sector ROA: 0.6 percent in 2017 and 0.5 percent in 2018.
- Bank profitability rebounded in 2017–18, driven by large and small banks; medium banks’ profitability lagged.
- Low interest margins and provisioning needs have dampened return on equity.

### Sovereign exposure and vulnerabilities
- Banks’ exposures to the domestic sovereign are over 11 percent of total assets (share of Domestic Government Bonds, as of November 2018: Italy 10.5).
- BdI estimate: an upward shift of 100 bps in the government yield curve would reduce banks’ CET1 ratio by 40 bps.
- Banks are moving a large share of sovereign bonds from fair-value to amortized-cost (AC) accounting and reducing duration of sovereign holdings.
- High concentration of sovereign debt increases vulnerability of banks’ capital and liquidity to adverse market valuations and affects equity prices and funding costs.

### Support measures and transitional arrangements
- Italian banks are the largest users of TLTRO at close to €250 billion, boosting liquidity and profitability.
- IFRS 9 transitional arrangements allowed banks to increase NPL provisions while deferring impact on capital; transitional arrangements will be gradually phased out by 2022.
- Estimated total IFRS 9 transitional impact: around 104 bps for SIs and 138 bps for LSIs.
- Transitional arrangements have enhanced banks’ ability to reduce NPLs through securitizations or sales, but further disposals may result in additional costs and could impact capital levels.

### Stress-testing objectives, scope, and key design features
- Objective: assess capacity of the banking system to withstand severe but plausible macroeconomic shocks and identify weaknesses, channels of propagation, and policy priorities.
- Stress tests covered solvency, liquidity, and contagion risks, plus profitability analysis and a non-financial corporate sector stress test.
- Solvency tests: three-year horizon from 2018:Q4 to 2021:Q4; used IMF satellite models and methodologies; included scenario-based and single-factor tests.
- Liquidity tests: LCR and NSFR regulatory approaches and a cashflow-based maturity ladder approach to assess resilience to large withdrawals.
- Contagion analysis: examined domestic interbank and inter-sectoral linkages and cross-border spillovers.
- Top-down (TD) stress tests used EBA ITS templates (FINREP and COREP) with end-2018 as starting point, complemented by BdI historical default data (BdI Credit Register), Bloomberg, Haver Analytics, Moody’s KMV, Fitch, and WEO.

### Main findings from TD stress tests and vulnerability analysis
- Sample: solvency stress test covered 9 SIs accounting for 68 percent of banking sector assets (the Italian banking sector includes 11 SIs accounting for about 74 percent; two SIs comprising about 6 percent were under restructuring and excluded).
- Solvency stress test result for the 9 SIs: total CET1 ratio declined by 370 bps from 11.9 percent to 8.2 percent.
- Under baseline scenario, 2 SIs would fall below minimum capital requirement thresholds; under the stress scenario, 3 banks (comprising about 8.5 percent of the banking system’s assets) fall below thresholds.
- Resulting capital shortfall against capital thresholds is about 0.2 percent of GDP.
- Capital needed to bring the CET1 ratio of the 9 SIs back to the end-2018 level of 12 percent is about 2.2 percent of GDP.
- Liquidity stress tests suggest relatively comfortable positions but are boosted by TLTRO use and have a high concentration of liquid assets in Italian government securities, increasing sovereign risk vulnerability.

### Adverse scenario design and macroeconomic impact
- Solvency stress test scenarios: baseline corresponds to April 2019 WEO projections (slowdown in real GDP growth and continued elevated sovereign yield levels). Adverse scenario simulated using the Global Macrofinancial Model (Vitek 2018).
- Adverse scenario main feature: reemergence of sovereign stresses in Italy, leading to a sharp rise in risk premia.
- Assumed shocks: 180 bps rise in Italy’s long-term risk premium (peak increase relative to baseline); term premia reduced by 70 bps in the EA core and by 40 bps in other advanced economies due to safe-haven inflows.
- Trigger: loss of investor confidence linked to concerns about rollback of reforms, less market-friendly policies, and expansionary budget proposals relative to European commitments.
- Macroeconomic effects: sharp drop in investment and output; recession worsens primary fiscal balance via lower tax receipts; tightening of fiscal position exacerbates downturn; corporates face higher borrowing rates and sharply reduce investment.
- Scenario impact on GDP: peak-to-trough decline in GDP throughout the horizon is 5.1 percent; represents a fall of 7.2 percent relative to baseline by 2021.

### Policy implications and recommendations
- Authorities should adopt measures to build further resilience, including enhancements to banks’ capital levels to ensure all banks maintain adequate capital ratios under stress scenarios.
- A thorough supervisory review of banks’ business models and governance is advisable, using the full gamut of the supervisory toolkit to achieve improvements, consolidation, or orderly winddowns as needed.
- Recognize limitations: FSAP stress test results differ from other institutions (EBA, ECB) due to different methodology, sample, macro scenarios, data input and parameters; tests were carried out in cooperation with ECB and BdI.
- Stress test results are macroprudential tools and should not be interpreted as macroeconomic forecasts; adverse scenarios represent tail events based on historical distribution and are subject to data and methodological limitations.

*Source: INTRODUCTION (1itaea2020008)*

### 6.9 percent fall in consumption and a 22.3 percent fall in investment. The deviation of growth

### 1itaea2020008 - 6.9 percent fall in consumption and a 22.3 percent fall in investment. The deviation of growth

### Adverse scenario: macroeconomic and financial impacts
- Consumption falls by 6.9 percent and investment falls by 22.3 percent (scenario headline).
- The deviation of growth relative to the baseline is equivalent to a 2.1 standard deviations shock to GDP growth in the second year of a three-year horizon (U-shaped profile).
- Long-term yields increase by 180 bps in the second year (relative to the baseline).
- Spreads relative to German long-term Bunds reach 580 bps in the second year of the scenario before narrowing to 560 bps.
- Heightened risk aversion reduces equity prices by 45 percent in Italy over two years, relative to the baseline.
- The shock is broadly consistent with a 5 percent growth-at-risk measure.
- Under the scenario, Italian sovereign yields reach up to 550 bps (noted as above the EBA 2017 stress test yield by 150 bps).

### Scenario calibration (Table 2 and Table 3 figures preserved)
- Deviation from the baseline (Adverse Scenario, in percentage points unless specified):
  - Real GDP: 2019 -3.4, 2020 -6.9, 2021 -7.2
  - Policy interest rates: 2019 0.1, 2020 0.1, 2021 0.0
  - Short-term money market rate: 2019 0.8, 2020 1.1, 2021 0.9
  - Spread of short-term money market rate: 2019 0.7, 2020 1.0, 2021 0.9
  - Long-term government bond yield: 2019 1.3, 2020 1.8, 2021 1.5
  - Real effective exchange rate appreciation (-)/depreciation (+): 2019 8.0, 2020 12.7, 2021 12.3
  - Nominal exchange rate appreciation (-)/depreciation (+): 2019 7.7, 2020 11.5, 2021 10.4
  - Inflation rate (CPI): 2019 -0.3, 2020 -1.2, 2021 -0.9
  - Unemployment rate: 2019 1.1, 2020 3.4, 2021 4.5
- Memo from Table 2:
  - Baseline Real GDP growth (in percent): 2019 0.0, 2020 0.9, 2021 0.7
  - (Adverse) Real GDP growth (in percent): 2019 -3.3, 2020 -2.7, 2021 0.3
  - Cumulative real GDP growth (from 2018): 2019 -3.3, 2020 -6.0, 2021 -5.7
  - Severity: deviation of growth from baseline in multiples of standard deviation of historical growth volatility: 2019 1.9, 2020 2.1, 2021 1.6
- FSAP ST Baseline and Adverse Scenario (Table 3):
  - Baseline (2019/2020/2021): Real GDP Growth 0.0 / 0.9 / 0.7; Policy interest rate 0.0 / 0.2 / 0.3; Short-term money market spread 0.0 / 0.0 / 0.0; Long-term government bond yield 3.5 / 3.8 / 4.0; Nominal exchange rate app (-)/dep(+) 0.0 / 0.0 / 0.0; Inflation rate (CPI) 0.8 / 1.4 / 1.6; Unemployment rate 10.8 / 10.6 / 10.4.
  - Adverse (2019/2020/2021): Real GDP Growth -3.4 / -2.7 / 0.3; Policy interest rate 0.1 / 0.3 / 0.3; Short-term money market spread 0.8 / 1.1 / 0.9; Long-term government bond yield 4.8 / 5.5 / 5.5; Nominal exchange rate app (-)/dep(+) 7.7 / 11.5 / 10.4; Inflation rate (CPI) 0.5 / 0.2 / 0.7; Unemployment rate 11.9 / 14.0 / 14.9.

### Methodology: stress testing framework and balance sheet projections
- The TD exercise for banks uses the IMF’s internally developed solvency stress testing framework, covering credit risk (all exposures), market risks (equity, exchange and interest rate risks), sovereign risk, and interest rate risk on the banking book. The derivatives book was not considered due to insufficient granular data.
- Balance sheet growth:
  - Quasi-static approach: asset allocation and funding composition held constant; balance sheets grow in line with the nominal GDP path specified in the stress test scenario.
  - A floor on the rate of change of balance sheets was set at zero percent to prevent deleveraging; this constraint was binding in the adverse scenario.
  - Balance sheet growth estimated per bank using weighted average GDP growth of countries with significant exposures.
  - Other influences: revaluation from FX movements and conversion of off-balance sheet items (credit lines and guarantees) to the balance sheet.
- Projection of RWAs:
  - Differentiated between standardized (STA) and internal ratings-based (IRB) portfolios.
  - For STA portfolios, RWAs change due to balance sheet growth, new provisions for credit losses, exchange rate movements, and the triggered portion of off-balance sheet items.
  - For IRB portfolios, projected through-the-cycle (TTC) PDs for each asset class/industry used to calculate new average risk weights.
  - EAD projection driven by balance sheet assumptions, structural FX risk in foreign currencies, and triggered portion of off-balance sheet items.
  - Triggered portion assumed to be 0 percent for the baseline and 2.5 percent for adverse scenarios.

### Credit risk: composition, PD estimation, and projections
- Credit risk accounted for 88 percent of total RWAs in the sample banks.
- Asset composition:
  - Loans represent 71 percent of assets, followed by debt securities.
  - Loans by sector: large firms 27 percent, mortgages 19 percent, SMEs 17 percent, other financial corporations 11 percent.
- IRB versus STA:
  - Out of nine banks, two apply exclusively STA; remaining banks apply IRB partially.
  - IRB approach accounts for 50 percent (median) of total credit exposure of the seven sample IRB banks, range 24 percent to 71 percent.
- Geographic distribution:
  - 60 percent of total loans are distributed in Italy.
  - For customer loans (non-financial corporations and households), 65 percent of loan exposure is in Italy; remaining exposures include Germany, Austria, and Turkey.
  - Three SIs have >30 percent of loan exposures outside Italy.
- Default rate estimation:
  - Historical default rates for five portfolios (corporate large firms, corporate SMEs, mortgage, consumer, financial institution) provided by BdI.
  - Point-in-time (PiT) PDs projected using regression models with macro variables; estimations used a Bayesian Model Averaging (BMA) approach.
  - PiT shifts applied to regulatory PDs for non-defaulted exposures.
  - For exposures outside Italy, Moody’s expected default frequency (EDF) used for significant geographies (Germany, Austria, Turkey, Russia, United States) and mapped to COREP portfolios.
- PD dynamics:
  - Given weak baseline outlook, PDs in most segments projected to gradually increase in baseline and worsen further in adverse scenario.
  - Impact is larger for large firms and SMEs than for mortgages and financial institutions.
  - Relevant drivers: real domestic output, unemployment rate, inflation, short-term and long-term interest rates (higher posterior inclusion probability and sizable long run multipliers).

### Loss given default (LGD), provisioning, and NPL treatment
- Bank-by-bank LGD rates used, informed by FSAP analysis of recovery rates for NPLs (UTP and bad loans).
- Provisioning rate calculated per bank, differentiating secured and unsecured loans for UTP and bad loans.
  - Secured loss rate median: 42 percent of total NPLs (range 38 percent to 46 percent).
  - Unsecured loss rate median: 65 percent of total NPLs (range 54 percent to 72 percent).
- PIT parameters used to compute loan loss provisioning across all exposures (on- and off-balance sheet), with off-balance items migrating to on-balance sheet.
- FSAP estimated additional provisioning of about €4.9 billion (6.4 percent of existing provisions or 0.44 percent of RWAs) for all SIs and LSIs, largely related to the UTP portfolio.
- Assumptions on existing NPLs seasoning over the three-year horizon:
  - 50 percent of the UTP portfolio will become bad loans.
  - 20 percent will return to performing status.
  - 30 percent will be closed in banks’ books as UTP.
- Interest revenue conservatively assumed to accrue only on performing exposures in both baseline and adverse scenarios.
- NPL disposals:
  - Exercise considered additional costs from market-based NPL disposals to reach NPL ratio targets disclosed for 2021 by a subset of banks.
  - NPL disposal loss rate is, on average, 14 percentage points higher than loss rate from internal workout.
  - Additional provisions from disposal losses calculated as the difference between market-based sale loss rates and internal workout loss rates.
- All scenarios use fully loaded IFRS 9 capital ratios (i.e., transitional arrangements for IFRS 9 not applied).

### Capital, hurdle rates, and failure conditions
- Hurdle rates differentiated between systemic (O-SII/G-SIB) and non-systemic banks (Table 4):
  - Total Capital ratio (total capital to RWAs): OSII 8 + OSII buffer; Other banks (Non-OSII) 8
  - Tier I Capital ratio: OSII 6 + OSII buffer; Other banks 6
  - Common Equity Tier I (CET1) ratio: OSII 4.5 + OSII buffer; Other banks 4.5
  - Leverage ratio (Tier 1 capital to total assets): 3 for both groups
- O-SII buffers range from 0.19 percent to 1.0 in 2021; phase-in during 2018–2021 accounted for.
- Final capital levels calculated on a fully loaded basis, including IFRS 9 implementation.
- Banks ending the stress-test horizon with capital levels or leverage ratio below relevant hurdle rates are considered to have failed the test.

### Key quantitative results and risk contributions (selected)
- Credit RWAs account for 88.1 percent of total RWAs; market risk 4.2 percent; operational risk 7.6 percent (Figure 6 RWA by Risk Factor).
- RWA allocation: Standardized vs IRB 46.8 percent vs 53.2 percent (Figure 6 RWA by IRB and Standardized).
- Asset composition (percent of total asset): Loan 71.0; Debt securities 17.3; Non-interest assets 11.6 (Figure 7 Asset Composition).
- Loan by category shares (percent, share of total loan): Large firms/Corporate SME/Mortgage/Consumer and Others—detailed sectoral shares reported in the source figures and text.

_Italic: Source: IMF staff calculations as presented in the provided content._

### 40.      Own-sovereign securities constitute the largest share of bank holdings of debt

### 40.      Own-sovereign securities constitute the largest share of bank holdings of debt

### Composition of debt securities holdings
- Own-sovereign securities represented 44 percent of total debt securities (including FV and AC categories) as of December 2018.
- Foreign-sovereign securities represented 34 percent of total debt securities.
- Securities issued by credit and financial institutions represented 19 percent of total debt securities.
- Own-sovereign securities were mostly booked at FV (either FVTPL or FVOCI) at 61 percent of total own-sovereign securities.
- The total share of securities booked at FV accounted for 69 percent of total.

### Duration and valuation treatment
- Average duration of FV own-sovereign securities: around 4 years.
- Average duration of own-sovereign securities in AC category: 6 years on average.
- Foreign-sovereign securities average durations: around 5 years (FV) and 6 years (AC).
- Shorter duration of sovereign securities at FV reduced the impact of sovereign yield changes on regulatory capital.
- Valuation change formula used in stress test (modified duration approach):
  - ∆ Valuation / Valuation = -MD. ∆y_MD

### Market risk stress testing scope and treatment
- Market risks considered: changes in interest rates, exchange rates and equity prices.
- Exclusions: amortized cost positions held in a hedge-accounting relationship, and hedge-accounting derivatives.
- Market risk treated as an add-on component with separate calibration consistent with the macroeconomic scenario; impacts both capital resources (via profit and loss or OCI) and capital requirements (RWAs).
- Market valuation losses for debt securities estimated using modified duration:
  - For every country and year, sovereign yield curves constructed by linear interpolation of short- and long-term interest rates as specified in macro scenarios.
  - Average duration of debt portfolio calculated for each bank using supervisory data (COREP).
  - Losses = size of bond portfolio × average duration × change in yields (with non-sovereign yields moving in line with sovereign yield plus a credit spread along the three-year horizon).

### Equity revaluation floor (trading intent)
- For equities held with trading intent, FV impact was subject to a floor using an approach similar to EBA 2018 stress test methodology:
  - ∆Eq=1.5∗(−0.20% ( 퐸퐸퐸퐸 퐿퐿퐿퐿퐿퐿퐿퐿 +퐸퐸퐸퐸 푆푆ℎ퐿퐿표표표표 ) )
  - VaR scaling factor set to the upper bound of 1.5; trading position includes FV of equity instruments (assets) and short positions in equity instruments (liabilities).

### Model and behavioral assumptions for income and rates
- Net interest income projected using maturity gap analysis, tracking repricing up to three-year stress horizon; assumption: bank does not change maturity profile over stress period.
- Funding rates projected using satellite models with BMA regression techniques; mapped to five segments: retail overnight deposit, retail term deposit, wholesale overnight, wholesale term deposits, and debt securities; projections attached to bank-specific starting points using end-2018 COREP data.
- Lending rates projected using satellite models with funding rate projection as input; mapped to three segments: corporate, mortgage, and consumer loans; projections attached to bank-specific starting points using end-2018 COREP data.
- Interest income from debt securities projected based on changes in the yield of the respective securities for both baseline and adverse scenarios.
- Portfolio-level modified duration approach used to measure gains/losses in FV portfolios; rebalancing of the portfolio was not allowed through the horizon. For AC securities, provisions adjusted according to changes in credit spreads.
- Income projection:
  - Includes all risk factors; gains or losses from other market positions (commodity and currency net open position) impacted via scenario evolution.
  - Remaining income-statement items grow in line with balance-sheet size; net fee and commission income and operational/administrative expenses projected similarly.
  - Under the adverse scenario, growth in non-interest income and expenses subject to a zero percent floor.
  - Extraordinary income and loss assumed not to recur.
  - Income tax set as 30 percent of income before tax.
- Dividend policy in stress test:
  - Dividends paid at 30 percent of current period net income after taxes by profit-making banks and only if in compliance with supervisory capital requirements.
  - Banks not allowed to issue new shares or make repurchases during the stress-test horizon.

### Interest Rate Risk in the Banking Book (IRRBB)
- Aggregate funding and lending rates estimated via BMA satellite models for retail and wholesale deposits, debt securities, household and corporate loans; model outputs mapped to bank-specific interest paths by attaching period changes of aggregate rates to bank-specific starting points.
- IRRBB template used to map projected rates to financial assets and liabilities by product and counterparty; includes maturity ladder for fixed-rate instruments and repricing dates for floating-rate instruments.
- Explanatory variables for interest-rate models include GDP growth, inflation, exchange rate, unemployment rate, and short- and long-term interest rates (Italy-specific, EA-, and U.S.-based indicators).
- Funding cost included as explanatory factor in lending rate projection to allow partial pass-through of rising funding cost to lending rates.
- PDs from credit risk satellite models included as add-on risk factor in lending rate determination for household and corporate segments.
- A conservative lower bound within lending-rate forecast confidence band selected: 5th percentile for the adverse and 25th percentile for the baseline.
- Resulting average decline in net interest margin: 0.07 percentage points under the baseline and 0.18 percentage points under the adverse scenario.

### Key drivers in interest income and funding cost projections
- Money market rates and long-term sovereign yields are main contributors.
- On funding side: 3-month money market rate and 10-year domestic sovereign bond yield spread explain most movement in interest expense.
- On lending side: similar variables plus U.S. long-term sovereign bond yield and funding cost associated with retail term deposits.
- Long-run pass-through: Italy’s sovereign bond yield and short-term interest rate significantly affect funding rates, particularly government bond yield → retail term deposit rate, and short-term rate → wholesale term deposit rate and retail/wholesale overnight deposit rates.

### Funding and lending pricing details
- Repricing rules:
  - Repos priced following short-term money market rates (EURIBOR); interest rate projections tied to baseline and adverse scenarios (adverse scenario: maximum 1.1 percentage points deviation from baseline).
  - Retail and wholesale deposit funding repriced according to satellite-model outputs across four deposit categories (wholesale overnight, wholesale term, household overnight, household term).
  - Debt securities pricing estimated by satellite models for new bonds with contractual maturities above one year.
- Interest income measured using effective lending rates weighted by December 2018 notional shares for consumer loans, mortgages and non-financial corporate loans.
- Rates on debt securities follow projections on long-term sovereign bond yield linked to macro scenario.

### Solvency stress test results (three-year horizon)
- Under the baseline scenario, FSAP LGD assumptions (including transition of 50 per cent of UTPs to bad loans) increased provisions needed.
- Aggregate CET1 ratio impacts:
  - With “no NPL disposal”: aggregate CET1 ratio declines by 56 bps from 11.9 percent to 11.4 percent (decline mostly attributable to increase in RWAs).
  - With “NPL disposal”: CET1 ratio declines by 102 bps to 10.9 percent.
- Capital threshold breaches:
  - With “no NPL disposal”: one bank falls below the T1 threshold and another bank falls below the CAR threshold.
  - With NPL disposal targets, the first bank would fall short of the three capital thresholds.
  - Resulting shortfall in capital is about 0.05 percent of GDP.
  - Two banks see leverage ratio fall below the minimum threshold of 3 percent.
- Main contributors to drop in capitalization under “NPL disposal”:
  - Credit risk (additional loan-loss provisioning) accounted for 4.5 percentage points of CET1 deterioration.
  - Increase in RWAs contributed 0.5 percentage point (mostly from positive asset growth).
  - Increase in sovereign yields impacted via valuation losses and NII reduction related to higher funding costs: 0.4 percentage points each.
- Pre-provision revenue (aggregate NII, non-interest income, and non-interest expenses) increased aggregate CET1 by 4.9 percentage points relative to the starting point.

*Source: IMF staff analysis as presented in the provided content.*

### 62.      The adverse scenario has a significant impact on banks’ capital ratios (Figure 12).

### 1itaea2020008 - 62.      The adverse scenario has a significant impact on banks’ capital ratios (Figure 12)

### Adverse scenario impact on Significant Institutions (SIs)
- Without the inclusion of NPL disposal plans:
  - Average fully loaded CET1 capital ratio declines from 11.9 percent in 2018 to 8.2 percent in 2021.
  - Three banks would see their capital ratios drop below the thresholds, with an average CET1 ratio decline of 9.3 percentage points.
  - The asset share of the three banks is 8.5 percent of the sample’s assets.
  - The resulting shortfall in capital is about 0.2 percent of GDP.
- Results do not change materially with the inclusion of NPL disposal plans.
- Aggregate leverage ratio:
  - Would decline by 1.4 percentage points from 5.8 percent to 4.4 percent.
  - The three failed banks would see their leverage ratio decline below the 3 percent threshold.
- Key drivers of CET1 declines:
  - Credit risk provisioning: about 5.3 percentage points decline in average CET1.
  - Valuation losses due to increase in sovereign yields: 1.2 percentage point decline in CET1.
  - NII reduction related to higher funding costs: 0.9 percentage points decline in CET1.
  - Expansion of RWAs (driven by heightened credit risk, PiT shifts, valuation effects in FX exposures, and projected loan paths) lowers CET1 by 1.2 percentage point across sample banks.
  - Pre-provision revenue (aggregate NII, non-interest income, non-interest expenses) increases aggregate CET1 by 4.7 percentage points relative to the starting point.
- Capital shortfall metrics:
  - Resulting capital shortfall against the threshold is small at about 0.2 percent of GDP.
  - Capital needs to bring the CET1 ratio of the 9 SIs in the stress scenario back to the end-2018 level of 12 percent is about 2.2 percent of the GDP.
- Sensitivity analysis:
  - FSAP increased loss rates by 20 percent (LGD increase) in a sensitivity run.
  - CET1 ratios were also evaluated against a higher threshold of 7 percent (incorporating the conservation buffer, CCB).
  - Number of bank failures does not change in the adverse scenario if a CET1 threshold of 7 percent was used or LGD rates were increased by 20 percent individually.
  - Combining CET1 threshold of 7 percent and a 20 percent increase in LGD: an additional bank will fall slightly below the 7 percent CET1 threshold.
- Specific table/figure summary data (SIs, selected scenarios):
  - Baseline - end-2018: Average CET1 ratio 11.9000005.6 (table formatting preserved in source).
  - Baseline - end-2021 Without NPL disposal plan: Average CET1 ratio 11.4011...0.015.5 (table formatting preserved).
  - Baseline - end-2021 With NPL disposal plan: Average CET1 ratio 10.9112...0.055.5.3 (table formatting preserved).
  - Adverse end of 2021 Without NPL disposal plan: Average CET1 ratio 8.22338.50.214.4 (table formatting preserved).
  - Adverse end of 2021 With NPL disposal plan: Average CET1 ratio 8.1 2338.50.254.3 (table formatting preserved).
  - Notes in table:
    - Leverage ratio proxied by Tier1 capital divided by Total assets (non-risk weighted).
    - Capital ratio is fully loaded IFRS9 implementation (post phase-in of initial IFRS 9 impact).

### Sensitivity analysis of Less Significant Institutions (LSIs) — overview and vulnerabilities
- Sector-wide metrics (LSIs, end-June 2018 reference):
  - CET1 ratio for the sector is 15.2 percent on a fully-loaded (FL) basis.
  - A quarter of banks by assets have fully loaded capital levels less than 10 percent.
  - NPL ratio for the group is 13 percent (22 percent for the corporate sector portfolio).
  - Some individual institutions have NPL ratios up to 35 percent.
  - Exposure to own-sovereign is 26 percent of assets for the group, up to 37 percent for smaller LSIs.
- Sample and tests:
  - Sample of 62 LSIs subjected to three static, one-factor sensitivity tests:
    - (i) sharp increase in yields (debt portfolio mark-to-market losses);
    - (ii) loan losses from new NPL formation;
    - (iii) IRRBB (interest rate risk in the banking book).
  - Capital ratios of about a quarter to one-third of the sample fall below the 7 percent CET1 threshold under tests (i) and (ii).
  - Impact of a sharp increase in yields is higher for LSIs than for SIs, reflecting higher concentration of Italian sovereign bonds in LSIs’ portfolios.
  - Reference dates: SIs December 2018; LSIs June 2018.

### A. Debt portfolios sensitivity test (method and scenarios)
- Coverage:
  - Applies to own-sovereign, financial and non-financial corporate, and securitized debt.
  - Valuation changes applied to FV portfolios via modified duration; provisions for AC portfolios using an LGD of 45 percent.
- Scenario calibration:
  - Compute daily 30-day change in yields since 2000; choose kth percentile of monthly increase in average of LT (10-year) and ST (3m) yield.
  - Three calibrations used (Table 5):
    - Scenario 1: 95th percentile of year 2011 — ST 3.2, LT 1.4, Average 2.3, Date 12/19/2011.
    - Scenario 2: 98th percentile, 2000–18 — ST 2.1, LT 0.4, Average 1.2, Date 10/14/2008.
    - Scenario 3: 95th percentile, 2000–18 — ST 0.8, LT 0.8, Average 0.8, Date 8/16/2012.
  - Scenario 1 and 2 assumed a flattening yield curve (ST shock > LT shock).

### Debt portfolio sensitivity test results (aggregate findings)
- LSIs:
  - Most stressful scenario (Scenario 1, 2011 rapid rise in yields):
    - Capital losses amount to 300 bps for the group of LSIs.
    - Almost a quarter of the group by assets see their CET1 ratio fall below 7 percent.
    - Total capital shortfall of 0.02 percent of GDP.
    - Estimated capital needs to bring sample banks’ CET1 back to starting ratio of 15.2 percent is 2.1 percent of GDP.
  - Scenario impacts (Table 6):
    - Aggregate CET1 impact (bps): SIs — 131, 75, 52 for Scenarios 1,2,3 respectively; LSIs — 300, 185, 120 for Scenarios 1,2,3 respectively.
    - <7% CET1 — #banks (SIs): 0,0,0; (LSIs): 10,3,2.
    - <7% CET1 — Share of assets: SIs 0%,0%,0%; LSIs 22%,6%,6%.
    - <4.5% CET1 — #banks: SIs 0,0,0; LSIs 4,1,0.
    - Losses (share of assets): SIs 0%,0%,0%; LSIs 1%,0%,0%.
- SIs:
  - Among the 9 SIs included in the test, no bank falls below the 7 percent threshold in the debt portfolio sensitivity tests.
- Notes:
  - Data are as of June 2018 for LSIs; in the second half of the year share of debt securities in FV portfolio declined as banks moved some FV portfolio to AC, so impact of sovereign shock would be lower than in simulations.

### B. Loan losses due to NPL flows (LSIs)
- Test scope:
  - Considers new flows of NPLs for household loans (HH), non-financial corporates (NFC), and government loans (GG).
  - LSIs subjected to flows in line with historically worst observed NPL flows.
  - Provisioning needs associated with new NPL flows set at 60 percent of stock of new NPLs.
  - Include provisions, if any, required to bring coverage ratio for existing stock of NPLs up to 60 percent.
- Calibration and scenarios:
  - Shocks defined based on historical quarterly PD percentiles since 2000, adjusted for LSIs’ higher PDs and bank-specific starting PDs.
  - Two scenarios:
    - Scenario 1: Maximum PDs for HH, NFC, GG since 2000, system-wide.
    - Scenario 2: 80th percentile PDs for HH, NFC, GG since 2000, system-wide.
  - As of June 2018, system-wide new flows are around the median of new flows observed since 2000s.
- Results:
  - Most severe scenario:
    - Capital losses amounted to 365 bps for the group of LSIs.
    - 14 banks (35 percent of the group by assets) fall below the 7 percent CET1 ratio.
    - Capital needs to bring banks’ CET1 ratio back to 7 percent are about 0.1 percent of GDP.
    - Estimated capital needs to bring sample banks’ CET1 back to starting ratio of 15.2 percent is 0.26 percent of GDP.
  - Table 7 summary (End-June 2018):
    - Aggregate CET1 impact (bps): Scenario 1 — 365; Scenario 2 — 311.
    - <7% CET1: # banks — 14 (Scenario 1), 12 (Scenario 2); share of assets — 35% (Scenario 1), 29% (Scenario 2).
    - Losses (share of assets): 1.9% (Scenario 1), 1.8% (Scenario 2).
    - <4.5% CET1: # banks — 6 (Scenario 1), 4 (Scenario 2); share of assets — 9% (Scenario 1), 8% (Scenario 2).
- Observations:
  - Credit losses are meaningful for some banks and larger in the corporate portfolio.
  - The provisioning assumption of 60 percent is conservative average: lower for mortgages, higher for consumer credit and some corporate credit.

### C. Interest Rate Risk in the Banking Book (IRRBB)
- Most LSIs have a positive interest-rate gap on their banking book, helping NII in case of increases in interest rates.
- BdI-conducted IRRBB exercise (results shared with FSAP team):
  - Average loss of economic value of equity due to a +/- 200 bps interest rate change is 2.6 percent.
  - A number of small banks face losses greater than 20 percent of economic value of capital in the case of a -200 bps interest rate shock; several come close.

### D. Concentration risk and SME supporting factor (SIs)
- SME supporting factor (CRR):
  - Factor value: 0.7619 deduction in capital requirements for exposures to SMEs.
  - FSAP assessed impact given large SME portfolio in Italy.
  - Assigned risk weights for assessment: 92 percent (average RW of SME exposure not subject to supporting factor) and 100 percent.
  - Results:
    - On average, SME supporting factor boosted CET1 ratio by 18 bps (assuming 92 percent RWA) or 27 bps (assuming 100 percent RWA).
    - Maximum impact on CET1 ratio is 60 bps for one sample bank (assuming 92 percent RWA).
- Concentration risk test:
  - Simultaneous hypothetical default of the five largest borrowers of each bank (net exposure).
  - Using zero recovery rate, simultaneous default of five largest borrowers would cause aggregate CET1 ratio of sample banks to decline by 3.4 percentage points.

*Source: IMF staff, from the FSAP chapter on Italy (figures, tables, and text as provided).*

### 11.9 percent to 8.5 percent. No banks would see their CET1 ratio below the 4.5 percent

### 1itaea2020008 - 11.9 percent to 8.5 percent. No banks would see their CET1 ratio below the 4.5 percent

### Credit sensitivity test: top-5 borrower defaults
- Using a provisioning rate of 100 percent, the simultaneous default of the five largest borrowers would cause the aggregate CET1 ratio of sample banks to decline by 3.4 percentage points from 11.9 percent to 8.5 percent. No banks would see their CET1 ratio below the 4.5 percent threshold.
- Using a provisioning rate of 65 percent, the simultaneous default of the five largest borrowers would cause the aggregate CET1 ratio of sample banks to decline by 2.2 percentage points from 11.9 percent to 9.7 percent. No banks would see their CET1 ratio below the 4.5 percent threshold.
- Data sources noted: FINREP, COREP, IMF staff estimates.

### Liquidity stress tests — summary and vulnerabilities
- Three liquidity stress frameworks were conducted: LCR, NSFR, and cashflow-based analyses for 11 SIs and 61 LSIs. Analyses conducted in EUR. The cashflow-based analysis simulates outflows over maturity buckets from 1 day to 90 days, versus the single 30-day window of the LCR.
- Two key liquidity vulnerabilities identified:
  - Aggregate liquidity is boosted by reliance on TLTRO: at close to € 250 bn and representing on average about 10 percent of banks’ total liabilities, Italian banks are the largest users of this facility in the EA.
  - Liquid assets are concentrated in own government securities, making liquidity ratios susceptible to adverse market valuations of these securities. Recommendation: diversify bond exposures to other non-Italy sovereign bonds, and diversify by maturity and type of asset.

### A. NSFR-based liquidity: structure and ratios
- Funding structure differences (data as of end-October 2018 for SIs; end-June 2018 for LSIs):
  - LSIs: Retail funding is the largest source of available stable funding (ASF) at 59 percent; corporate funding 20 percent; debt instruments 9 percent.
  - SIs: Retail funding represents 28 percent of total funding. SIs’ funding also includes central bank funding 13 percent, unsecured funding from financial institutions 17 percent, and unstable retail funding 17 percent.
- NSFR outcomes:
  - NSFR ratios are above 1 for the majority of SIs and LSIs.
  - SIs’ ASF has a lower factor compared to the European bank average due to reliance on less-stable and wholesale funding.
  - LSIs show a lower RSF factor compared to European banks’ average due to a large share of government bonds on their asset side, contributing to higher NSFR ratios.

### B. LCR-based liquidity stress test: scenarios and results
- LCR stress scenarios considered (beyond Basel CRR parameters):
  - Retail scenario key assumptions:
    - 10 percent run-off rates for stable retail deposits and 15–20 percent for less stable retail;
    - 75 percent (40 percent) run-off rates for operational (non-operational) deposits not covered by Deposit Guarantee Scheme;
    - 10 percent haircut on government securities for HQLA calculation.
  - Wholesale scenario key assumptions:
    - 100 percent run-off rates for wholesale funding from other financial institutions;
    - 50 percent run-off rates for operational and non-operational deposits not covered by Deposit Guarantee Scheme;
    - 10 percent haircut on government securities for HQLA calculation.
- Aggregate LCR levels (Table 8):
  - System-wide LCR (SIs): 156%
  - SIs under Retail Scenario: 92%
  - SIs under Wholesale Scenario: 110%
  - System-wide LCR (LSIs): 229%
  - LSIs under Retail Scenario: 157%
  - LSIs under Wholesale Scenario: 189%
- Banks falling below 100% LCR:
  - SIs: Number of banks with LCR<100% — LCR baseline 0; Retail scenario 9; Wholesale scenario 1.
    - Share of group assets: baseline 0%; Retail scenario 91%; Wholesale scenario 7%.
    - Liquidity shortfall (millions of euros): baseline 0; Retail scenario 16,768; Wholesale scenario 1,494.
    - Liquidity shortfall (share of liabilities of failed banks): baseline 0.0%; Retail scenario 0.9%; Wholesale scenario 1.0%.
  - LSIs: Number of banks with LCR<100% — baseline 1; Retail scenario 10; Wholesale scenario 7.
    - Share of group assets: baseline 0.5%; Retail scenario 26%; Wholesale scenario 16%.
    - Liquidity shortfall (millions of euros): baseline 10; Retail scenario 1,341; Wholesale scenario 433.
    - Liquidity shortfall (share of liabilities of failed banks): baseline 1.0%; Retail scenario 2.3%; Wholesale scenario 1.2%.
- Key interpretation:
  - Aggregate LCR is comfortable (160 percent for SIs and 230 percent for LSIs) but banks are particularly vulnerable to a retail event: in a severe retail scenario, 9 SIs accounting for 91 percent of the group’s assets fall below 100 percent LCR, though the system-average stressed LCR falls just below 100 percent and the liquidity shortfall represents only 1 percent of liabilities.

### C. Cashflow-based liquidity stress test (CFLST)
- Sample and purpose:
  - CFLST analyzes liquidity risk exposure and risk-bearing capacity for 11 SIs and 55 LSIs.
  - Focuses on net-funding gap (NFG), cumulated net-funding gap (CNFG), counterbalancing capacity (CBC), and cumulated CBC.
  - Uses COREP templates (Maturity Ladder, C66.00).
- Composition of CBC:
  - LSIs: 82 percent of CBC is composed of general government items.
  - SIs: government items make 67 percent of CBC.
  - Implication: high quality of CBC composition but higher sovereign market liquidity risk concentration.
- Asset encumbrance (AE) ratios:
  - Average AE SIs: 30 percent.
  - Average AE LSIs: 22 percent.
  - Some LSIs have AE ratios above 40 percent, indicating potential constraints on tapping unsecured wholesale funding and susceptibility to funding shocks.
- Contractual liquidity risk exposure (SIs sample):
  - Contractual outflows within the first month amount to about 75 percent of total assets.
  - Contractual inflows within the first month amount to about 38 percent of total assets.
  - The cumulative net funding gap over the first 4 weeks reaches about 37 percent of total assets or 818 billion EUR.
  - First-month outflow composition: 31 percent from retail deposits, 10 percent from corporate deposits, repos collateralized with 0 percent RW securities around 12 percent of contractual outflows in the first month.
  - First-month inflows drivers: reverse repos against 0 percent RW securities 15 percent, corporate and other inflows 10 percent combined.
- LSIs CFLST findings:
  - Retail deposits represent 41 percent of contractual outflows (37 percent in the first month).
  - Overnight corporate deposit exposure 12 percent of contractual outflows; other deposits 8 percent.
  - Repos against 0 percent RW securities: 16 percent of contractual outflows in the first month.
  - Contractual inflows in first month: retail, corporate, central bank inflows around 9 percent of total inflows in the first month (versus 48 percent for entire horizon).
  - Reverse repos against 0 percent RW securities about 6.5 percent of contractual inflows in the first month.
- Run-off calibration (Table 9): range of run-off factors (in percent) across mild, medium, severe and most severe scenarios
  - Unsecured LT/ST Issuances and financial deposits: 30–100
  - Secured issuances: 30–100
  - Stable retail deposits: 0.5–10.5
  - Unstable retail deposits: 1.25–29.3
  - Operational corporate deposits (NFCs): 1.25–38
  - Non-operational corporate deposits (NFCs): 20–55.7
  - Repo across all collateral classes*: 100
  - Deposits outflows from financial institutions: 50–100
  - FX-Swaps in-/outflows: 15–100
  - Derivative in-/outflows: 25–100
  - Retail / corporate inflows: 0
  - Central bank inflows: 100
  - Other entities inflows: 0–30
  - Financial institutions inflows: 30–100
  - Other inflows: 0-100
  - Committed lines provided by the bank (FI): 50–100
  - Committed lines provided by the bank (non-FI): 50–100
  - Margin calls (derivatives): 2–10
  - Adverse market outflows (derivatives, HLBA): 0–100
  - Outflows due to rating downgrades: 0–100

### Policy-relevant observations and recommendations (as implied by analysis)
- Liquidity diversification:
  - Reduce concentration of liquid assets in own government securities by diversifying bond exposures to non-Italy sovereign bonds and varying maturities and asset types.
- Funding-source resilience:
  - Reduce reliance on TLTRO (currently close to € 250 bn and about 10 percent of banks’ total liabilities) and on unstable wholesale funding, particularly for SIs.
- Manage asset encumbrance:
  - Address high AE ratios, especially for LSIs with AE above 40 percent, to preserve ability to access unsecured wholesale markets and central bank liquidity.
- Enhance retail-liquidity resilience:
  - Given vulnerability to retail deposit runs (retail scenario causes many SIs to fall below 100 percent LCR), strengthen measures to shore up retail deposit stability and contingency liquidity buffers.

*Source: IMF staff calculations and analysis as contained in the provided content.*

### 89.      The cashflow liquidity stress test runs a large set of embedded scenarios of increasing

### 1itaea2020008 - 89.      The cashflow liquidity stress test runs a large set of embedded scenarios of increasing

### Stress test design and scenarios
- Time horizons considered: 4-week, 3 months, and 5-day.
- For each time-horizon: a baseline scenario and four stress scenarios with increasing severity: mild market stress, medium market stress, severe market stress, and a most severe market stress.
- Three approaches to counterbalancing capacity (CBC):
  - Full CBC: fully endogenous liquidity supply by the central bank as long as banks have unencumbered eligible collateral.
  - Full CBC with market haircuts: a full CBC is assumed, but market-specific haircuts and bank-specific market price effects are imposed on elements of the CBC. (Specific haircuts calibrated based on discussions with the authorities and the ECB.)
  - Partial CBC with market haircuts: non-marketable components of the CBC (i.e., credit claims and committed lines provided to the banks) are disregarded; market haircuts are imposed.

### Cashflow liquidity stress test — One-month (4-week) horizon: key findings
- SIs (significant institutions):
  - SIs are resilient in the face of a four-week shock; liquidity remains ample in the system.
  - Average scenario impact on SIs can range from -3 percent of total assets in a mild market to -12 percent of total assets in a most severe market.
  - Total CBC as a share of SIs’ total assets can decline from 15 percent to 5 percent across scenarios.
  - Potential liquidity needs could reach 12 billion euros in the most severe market scenario.
  - Table 10 (one-month horizon, Scenario 1): Remaining CBC as a share of total Assets for SIs: 15% (Mild Market), 12% (Medium Market), 10% (Severe Market), 8% (Most Severe Market).
  - Table 10 (one-month horizon, Scenario 2): Remaining CBC as a share of total Assets for SIs: 15%, 10%, 7%, 5%.
  - Table 10 (one-month horizon, Scenario 3): Remaining CBC as a share of total Assets for SIs: 13%, 10%, 7%, 5%.
  - Number of SIs unable to maintain a positive CBC in Table 10:
    - Scenario 1: 0, 0, 0, 1 (Mild, Medium, Severe, Most Severe).
    - Scenario 2: 0, 0, 0, 2.
    - Scenario 3: 0, 0, 0, 3.
- LSIs (less significant institutions):
  - Some LSIs are less resilient; potential liquidity needs become prevalent even in a mild market scenario for a few LSIs.
  - LSIs face CBC shortfalls ranging from 20 million euros in a medium stress market scenario to almost 1.1 billion euros in the most severe market scenario.
  - Scenario impact for LSIs can range from -3 percent to -13 percent of LSIs’ total assets.
  - Table 10 (one-month horizon, Scenario 1) Remaining CBC as a share of total Assets for LSIs: 21%, 20%, 20%, 18%.
  - Table 10 (one-month horizon, Scenario 2) Remaining CBC as a share of total Assets for LSIs: 17%, 15%, 14%, 12%.
  - Table 10 (one-month horizon, Scenario 3) Remaining CBC as a share of total Assets for LSIs: 16%, 14%, 13%, 11%.
  - Number of LSIs unable to maintain a positive CBC in Table 10:
    - Scenario 1: 0, 2, 2, 4.
    - Scenario 2: 0, 3, 7, 11.
    - Scenario 3: 1, 4, 7, 15.

### Cashflow liquidity stress test — Three-month horizon: key findings
- SIs:
  - On aggregate, SI banks are moderately resilient over the three-month horizon, but stress can be substantial.
  - Under the most severe market scenario, the average impact can amount to -18 percent of SIs’ total assets.
  - Potential liquidity needs for SIs could reach up to 100 billion euros in the most severe market scenario.
  - Table 11 (three months, Scenario 1) Remaining CBC as a share of total Assets for SIs: 12%, 10%, 5%, 2%.
  - Table 11 (three months, Scenario 2) Remaining CBC as a share of total Assets for SIs: 8%, 5%, 3%, 1%.
  - Table 11 (three months, Scenario 3) Remaining CBC as a share of total Assets for SIs: 10%, 7%, 2%, -2%.
  - Number of SIs unable to maintain a positive CBC in Table 11:
    - Scenario 1: 0, 0, 2, 5.
    - Scenario 2: 0, 0, 3, 6.
    - Scenario 3: 0, 0, 6, 8.
- LSIs:
  - LSIs face material liquidity strains over three months; potential liquidity needs can arise early.
  - Potential liquidity needs for LSIs can arise in a mild market scenario (−1.4 million euros) and reach almost 11 billion euros in the most severe market scenario.
  - 32 out of the 55 LSIs in the sample are not able to maintain a positive CBC in case of a most severe market stress.
  - Table 11 (three months, Scenario 1) Remaining CBC as a share of total Assets for LSIs: 20%, 19%, 14%, 10%.
  - Table 11 (three months, Scenario 2) Remaining CBC as a share of total Assets for LSIs: 16%, 14%, 8%, 4%.
  - Table 11 (three months, Scenario 3) Remaining CBC as a share of total Assets for LSIs: 15%, 13%, 7%, 3%.
  - Number of LSIs unable to maintain a positive CBC in Table 11:
    - Scenario 1: 3, 4, 14, 23.
    - Scenario 2: 3, 6, 25, 30.
    - Scenario 3: 4, 8, 27, 32.

### Cashflow liquidity stress test — Five-day horizon: key findings
- All SI and LSI banks can maintain a positive CBC during a five-day stress horizon.
- CBC as a share of total assets in the most severe scenario:
  - SIs: declines to 7 percent of total assets.
  - LSIs: remains as high as 17 percent of total assets.

### Cross-cutting conclusions from the liquidity analysis
- Banks are resilient to liquidity shocks in the short term but less so in a prolonged stress environment.
- Most SIs and LSIs remain liquid at the one-month horizon, even in the severe scenario, but materially less so at the three-month horizon.
- Banks that feature a negative CBC after stress are heterogeneous in causes:
  - Some have low initial CBCs.
  - Others have relatively large shares of credit claims in their initial CBC.
  - Some face high outflows due to committed lines to customers; others due to outflows from deposits of financial institutions or other deposits.

### Profitability analysis — drivers and system overview
- Dataset: 381 banks from S&P’s SNL database, covering 2010–2017; includes 99 joint stock (SpA) companies, 26 mutual (“popolari”) banks, and 254 credit cooperative banks (BCCs).
- Bank size groups: largest (>30 bn euros), mid-sized (4 bn euros to 30 bn euros), smallest (<4 bn euros).
- System-wide profitability:
  - Profitability rebounded in 2017–18 to levels close to the EU median, driven largely by large banks.
  - Medium and small banks’ profitability remained below the EU median despite improvements since the twin crises.
- Structural and cyclical headwinds:
  - Small- and mid-sized banks retain high operating costs compared to large banks and the EU median.
  - Cyclical factor: low and negative interest rates reduced net interest income; net interest income as a share of operating income declined significantly for all banks from 2010 to 2017.
- Decomposition of changes in ROA between 2010 and 2018:
  - Systemwide improvement driven mainly by large banks.
  - Major negative contributor: net interest income (NII) for the banking sector, large-sized banks, and small-sized banks.
  - Medium-sized banks show a positive contribution from NII between 2010 and 2018.
  - All banks benefited from increases in non-interest income.
  - Medium-sized banks’ profitability negatively impacted by increased provisioning needs.
  - Large banks reduced operating expenses; medium and small banks saw operating expenses increase on net.

### Impact of TLTRO on banks’ profitability
- Italy’s banks are main users of ECB TLTROs: about 200 billion euros of the ECB’s TLTROs allotted to Italy’s SIs.
- Banks benefit from a negative interest rate of -0.4 percent on TLTRO II borrowings (estimated at about 2 percent of SIs’ NII on average).
- Replacing TLTRO funding with market-based sources will be costlier for banks.
- TLTRO-III announced in March 2019: new series starting in September 2019 and ending in March 2021, each with a maturity of two years; will continue to support the banking system in the near term.
- Market cost of equivalent funding depends on funding mix and market conditions; short-term funding while maintaining comfortable LCRs/NSFRs could incur relatively low funding costs (e.g., ECB MRO at 0.0 percent, retail deposits at 0.38 percent).
- Banks’ CDS spreads rose 25 percent on average after May 2018; more recently CDS spreads have declined.
- Selected bank bond issuance spreads (May 2018–January 2019):
  - UniCredit 11/27/2018 USD 3 billion 12/4/2023 Sr Non-Preferred 420 bps.
  - UniCredit 1/15/2019 USD 2.5 billion 1/14/2022 Sr Non-Preferred 365 bps.
  - Intesa 8/30/2018 EUR 1.1 billion 8/30/2023 Sr Preferred 188 bps.
  - MPS 1/29/2019 EUR 1.1 billion 1/29/2024 Secured 190 bps.
  - Bond issuance prior to May 2018 examples: UniCredit 1/18/2018 EUR 1.8 billion 1/18/2023 Sr Non-Preferred 70 bps; UniCredit 4/12/2017 USD 1.2 billion 4/12/2022 Sr Preferred 200 bps (IMF staff roughly estimate ~190 bps via the 5yr EUR mid-swap rate for this item); Intesa 3/20/2018 EUR 1.5 billion 3/20/2028 Sr Preferred 77 bps.
- Illustrative estimates of annual market-based cost of an equivalent TLTRO funding (as of end-2018) — funding cost (percent) and impacts:
  - Retail deposit rate: Funding Cost 0.4; Gross TLTRO Replacement 0.5; TLTRO, Net of 50% of Banks’ Excess Reserves 0.4.
  - ECB’s MRO: Funding Cost 0.0; Gross TLTRO Replacement 0.0; TLTRO, Net of 50%... 0.0.
  - Long term bond – low volatility period (Intesa bond issuance, 2018): Funding Cost 1.8; Gross 2.2; Net 1.8.
  - Long term bond – high volatility period (UniCredit bond issuance, 2018): Funding Cost 4.2; Gross 5.2; Net 4.2.
  - Funding Mix 1: 30% Retail, 30% MRO, 40% Bonds (Intesa price): Gross 1.3; Net 1.0.
  - Funding Mix 2: 30% Retail, 30% MRO, 40% Bonds (UniCredit price): Gross 2.7; Net 2.2.
  - Funding Mix 3: 50% MRO, 50% Bonds (Intesa price): Gross 1.1; Net 0.9.
  - Funding Mix 4: 50% MRO, 50% Bonds (Unicredit price): Gross 2.6; Net 2.1.
- Note on estimates: The "Net TLTRO, Net of 50% of Banks’ Excess Reserves" assumes banks will replace TLTRO liquidity by using half of their excess reserves and financing the rest.

*Source: IMF staff calculations and analysis as presented in the provided content.*

### 107.      Costs of replacing TLTRO by market funding can have a meaningful impact on banks’

### 1itaea2020008 - 107.      Costs of replacing TLTRO by market funding can have a meaningful impact on banks’

### Replacement of TLTRO: potential funding-cost shock and bank equity impact
- Scenario setup:
  - If banks replace TLTRO by five-year bond issuance, issuance price could range from “188 bps + mid-swap rate” to “420 bps + mid-swap rate”.
  - Alternative scenario: banks can use up to half of their excess reserves at the ECB for their financing needs (“net TLTRO” scenario).
- Estimated impact:
  - If banks use funding mixtures, replacement costs could vary from 1–3 percent of banks’ equity on aggregate.
  - The impact varies significantly by bank.
- Additional considerations (not modeled here):
  - The cost from TLTRO refinancing could be passed-through to lending rates.
  - The cost could be mitigated by deleveraging.

### Market indicators and funding-cost drivers (selected observations)
- Banks’ CDS spreads increased in the second half of 2018 by about 25 percent, raising funding costs, but CDS spreads declined significantly since then.
- Cost of replacing TLTRO is costliest if banks choose long-term bond issuance.

### Corporate stress-test rationale and high-level findings
- Heightened sovereign spreads and slowing growth raise risks of adverse sovereign-bank-corporate feedback loops:
  - Higher sovereign yields increase marginal funding costs for Italian banks highly exposed to the sovereign, potentially leading to higher retail lending rates to corporates.
  - High NPLs can lead to credit rationing with negative growth effects.
- Purpose of the corporate stress test:
  - Assess the financial resilience of Italian corporates to adverse macroeconomic conditions using a micro-simulation model capturing interlinkages between macro environment and corporate vulnerability indicators.
  - Calibrated to the adverse macroeconomic scenario, the stress test reveals relatively high sensitivity of profitability, interest coverage ratio, and the share of debt held by vulnerable corporates to interest rate, profit, and combined shocks.

### Aggregate non-financial corporate (NFC) sector developments and vulnerabilities
- Households and macro context:
  - Households are substantial net lenders and the ‘richest’ institutional sector in Italy.
  - Household accumulated net financial assets of about 200 percent of GDP by end-2007, up from 156 percent of GDP at end-1995.
  - Households’ bad debt to banks account for about one-fifth of the economy-wide bad debt, or just above 5 percent of bank loans to households.
  - Share of vulnerable households in recent years has been around 11–12 percent, down from a peak of about 19 percent during the 2012 recession.
- NFC-level indicators:
  - Non-financial corporate sector net financial assets at -112 percent of GDP (slightly more indebted than the euro area average).
  - The corporate sector gross debt to surplus ratio is close to the euro area average.
  - Corporate indebtedness imposes a drag on total factor productivity growth.
  - Gross National Income (GNI) of NFCs before interest and taxes has declined and remains about 7 percentage points below the euro area average as a share of gross value added.
  - Unit labor cost (ULC) gap with European peers is over 30 percent.
  - Corporate profitability and net investment remain low.
- Historical improvements and remaining sensitivities:
  - Debt-servicing capacity improved due to monetary accommodation: interest expenses as a share of value added fell from around 8–10 percent pre-crisis to below 2 percent in 2017.
  - Ratio of GNI before interest and taxes to interest expense reached its historical peak of 13.5 per cent in 2017 (compared to an average of about 5 percent between 2002 and 2007).
  - Despite improvements, important segments remain vulnerable, contributing to higher debt-at-risk and posing risks for banks’ asset quality.

### Micro-data coverage and firm-level structure (CERVED database)
- Dataset overview:
  - The firm-level database compiled by CERVED covers 2002–17 and comprises more than 10 million observations for 1.5 million firms.
  - Sample coverage by firm size (2002-2017):
    - Total10,214,8451,531,853
    - Micro8,682,3401,454,908
    - Small1,127,716234,468
    - Medium327,11055,853
    - Large77,67911,335
  - About 80 percent of firms in 2017 are micro-firms with less than 10 employees; 1 percent are large firms with more than 249 employees.
  - Micro-enterprises account for about 28 percent of value added of the total business economy (Structural Business Survey, Eurostat).
- Data handling and limitations:
  - Database standardizes accounting presentations across industries.
  - Firms with simplified accounting (mostly micro firms) may not report full liability structure; this can bias sub-component analyses but not aggregate debt analyses.
  - Data cleaning steps include dropping firms with zero/missing total assets, tangible fixed assets, or sales; removing extreme year-to-year asset changes; addressing zero interest expense reports when financial debt is positive.
- Recent micro-level trends:
  - Over 2015–17, liquidity positions improved markedly and firms reduced leverage; gross debt stock close to 2007 levels and equity was built up.
  - Profitability remains structurally low: in 2017 both average and median EBIT over assets remain about 6-7 percent below average pre-crisis levels and about 17-18 percent below the 2007 peaks.
  - Median values are worse than averages across indicators, indicating a small set of high-performing firms and a long tail of weaker firms.

### Debt overhang metrics and debt-at-risk
- Vulnerability indicator:
  - Interest coverage ratio (ICR) defined as EBIT to interest expenses; ICR < 1 implies firms are unable to service debt under the status quo.
  - “Debt-at-risk” defined as the share of debt of firms with ICR less than 1.
- Debt composition for stress scenarios:
  - Stressed debt excludes advances, trade credits and pension and other fund liabilities, and accounts for about 65 percent of total debt.
- Recent improvements in debt servicing capacity:
  - Median ICRs almost doubled compared to pre-crisis years.
  - ICRs are lower for smaller firms and higher for larger firms; the gap has increased in recent years.
  - Share of firms-at-risk declined from around 40 percent at the crisis peak to 24 percent.
  - Debt-at-risk declined from around 35 percent at the crisis peak to 20 percent.

*Source: IMF staff analysis in 1itaea2020008 - chapter text provided.*

### 121. At the same time share of vulnerable firms in Italy as well as debt-at-risk remain

### 121. At the same time share of vulnerable firms in Italy as well as debt-at-risk remain strikingly high

### Vulnerability levels and profitability measures
- Share of vulnerable firms with ICR less than 1 declined from a pre-crisis average of 37 percent to 26 percent, though remains substantial.
- More than a quarter of firms are not able to generate sufficient operating income to cover their debt servicing costs.
- The share of debt held by firms with ICR less than 1 is 23 percent, increasing to 47 percent when also including companies with subpar performance (ICR between 1 and 3, following the categorization in IMF, 2019).
- Vulnerability indicators are high also for large Italian companies:
  - Out of all large companies 14 percent have ICR less than 1.
  - Those large companies with ICR less than 1 hold 18 percent of all debt of large companies and 40 percent of all corporate debt-at-risk.
- Recommendation on profitability metric for stress-tests: EBIT should be the preferred measure of profitability as it makes an explicit allowance for capital expenditures and may be more informative for comparing companies across different industries and capital intensities; EBITDA can make especially capital-intensive companies look like they have more funds to make interest payments.

### Distribution by firm size and debt concentration
- Gross debt and debt-at-risk are concentrated in micro and large companies.
- Micro enterprises are the most vulnerable category by most metrics, but historically the lion’s share of debt has been held by large companies.
- Trends:
  - Number of micro companies and their debt has been declining since 2012, reflecting higher net exit rates among more vulnerable micro firms.
  - Number of large firms rapidly recovered from the crisis impact, reaching their peak in 2017.
- Shares in 2017:
  - Increasing share of debt held by large companies reached 51 percent.
  - Large companies held 40 percent of debt-at-risk.
- Distributional characteristics:
  - Among micro firms debt-at-risk is rather uniformly distributed with a ratio of debt-at-risk to the share of vulnerable companies (ICR<1) close to 1.
  - Among large companies this ratio is close to 3, indicating debt-at-risk is concentrated in a smaller number of companies.
  - Overall, about 73 percent of debt-at-risk is held by micro and large companies.

### Sectoral concentration of vulnerabilities
- Vulnerabilities concentrated in four larger sectors that account for about 70 percent of debt-at-risk.
- Sectoral highlights (median ICR and share of debt-at-risk, 2017):
  - Manufacturing: more than one quarter of debt and debt-at-risk; relatively high median ICR but wide distribution of debt.
  - Construction and real estate: most vulnerable sectors with high shares of debt-at-risk and relatively low ICRs; together account for about 45 percent of gross firms’ non-performing exposures in mid-2018 (Banca d’Italia, 2018a).
  - Trade: accounts for another 14 percent of debt-at-risk; more than one-fifth of all micro-companies are concentrated in trade.
  - Professional services: close to 90 percent comprised of micro companies but one of the least vulnerable sectors (high ICR); protected by high entry regulations per OECD (2013).
- Sector table (selected entries, 2017): Total median ICR 3.3; Manufacturing ICR 4.6; Construction ICR 2.8; Real estate ICR 2.0; Trade ICR 3.7; Professionals ICR 5.0.

### Geographic concentration
- Debt and debt-at-risk are geographically concentrated.
- Northern areas have higher median ICRs compared to country-wide median values.
- North-west:
  - Almost one-third of total firms and one-third of all vulnerable firms are located in the north-west.
  - Firms in the north-west hold about 36 percent of all corporate debt in Italy.
  - Almost half of all debt-at-risk held by vulnerable firms in Italy is concentrated in firms in north-west.
- South:
  - Companies in the south hold 10 percent of overall corporate sector debt-at-risk.

### Debt composition and sensitivity to interest rates
- About 43 percent of total debt is held by firms with subpar ICR (less than 3, following IMF, 2019).
- More than 60 percent of total debt is held by companies with ICR less than 5.
- Improvement relative to 2012 sovereign crisis:
  - In 2012 as much as 65 percent of debt was held by firms with ICR less than 3; 76 percent for firms with ICR less than 5.
- Under historically low interest rate environment, the impact of a 100 bps interest rate shock on ICRs is proportionately higher than under more normal interest rate conditions.
- Assumptions on interest-rate sensitivity in stress-test:
  - About 35 percent of total debt are assumed less sensitive to interest rate developments (these comprise short-term trade credits, accounts payable, and some pension and other fund liabilities).
  - The remaining “effective debt” is sensitive and matches observed pass-through from new to outstanding lending rates.
  - Given average maturity of bank loans to NFCs in Italy assessed in the range of 4–5 years, assume about one-third of long-term debt will be repriced over the 3-year horizon of the stress test.

### Stress-test specification and calibration
- Three types of shocks applied to each firm: a profit shock, an interest rate shock, and a common shock; combined shock is simultaneous application of profit and interest rate shocks.
- Profit shock:
  - Determined econometrically by relating change in EBIT for every firm to GDP growth and other controls:
    ∆EBITijkt = α + β1 ∆ln(VAjt) + β2 Ukt + γ1j dj + γ2K dK + γt dt + eijkt
  - Regression panel across 21 economic sectors at NACE-2 level, 20 geographical regions, over 2004–17.
  - Estimated output semi-elasticity for the full sample is 0.84, though it varies considerably across firm size.
- Interest rate shock:
  - Related to ECB main refinancing rate and 10-year sovereign spreads over German bunds.
  - Calibration based on Albertazzi and others (2014): full long-run pass-through of ECB main refinancing rate to new lending rates.
  - Estimated pass-through of a 100 bps increase in sovereign spreads is close to two-thirds on average; in sovereign crisis scenarios pass-through to new lending rates in the long-term is estimated to be full.
  - Allow for more muted pass-through of sovereign spreads in the order of two-thirds given funding composition shifts.
- Combined shock:
  - Simultaneous application of profit and interest rate shocks; effects can reinforce each other and are not equal to the sum of individual shocks when implemented separately.

### Back-testing and model performance
- Back-testing constructed on 2012–13 sovereign crisis episode using 2011 as last observable data-point and out-of-sample macroeconomic data for 2012–13.
- Test results:
  - Model predicts reasonably well corporate vulnerability dynamics, share of vulnerable firms, and debt-at-risk at end of the 2012–13 test-period.
  - Model captures variations driven by domestic shocks (e.g., sovereign spreads).
  - Distributional performance:
    - Replicates very closely distributions of firms and debt by ICR categories, firms and “firms-at-risk” by firm size, and overall debt by firm size.
    - Debt-at-risk is marginally (by 0.8 pp) under-estimated while debt held by firms with ICR between 1 and 2 is slightly over-estimated.
    - Slight under-estimation of debt-at-risk held by large companies and over-estimation by 2 percentage points of debt held by small and medium sized firms.
    - Model predicts a median ICR of 40 and debt-at-risk of 35 percent at the 2012 turning point.
  - For the 2008–09 GFC episode the model captures turning points but does not replicate the full depth of the recession due to lower model-predicted decline in corporate profitability between 2007 and 2009.

_Italic: Source: CERVED; and IMF staff calculations (excerpts from 1itaea2020008)._

### 134.      The stress-test assesses firms’ debt servicing capacity under the FSAP adverse

### The stress-test assesses firms’ debt servicing capacity under the FSAP adverse

### Stress-test design and scenario assumptions
- The stress-test is conducted as an out-of-sample simulation exercise, covering a 3-year horizon.
- Baseline assumption: the outstanding amount of debt held by companies at the end of 2017 would be kept constant throughout the 3-year period of the stress test.
- Shocks calibration:
  - Shocks are calibrated to the 3-year cumulative difference in nominal GDP between the adverse scenario and the baseline used for the banking sector stress tests.
  - Sovereign spread shock is calibrated to the mean difference between Italian and German 10-year government bond yields over the three-year horizon.
- Scenario features:
  - The scenario is driven by Italy-specific shocks and does not assume any monetary accommodation.
  - Timing: the sovereign spread shock occurs almost fully in the first year, which could imply fast passthrough and repricing of outstanding debt over the stress-test horizon, especially if the shock is permanent in expectations.
- Horizon: 3 years; debt stock reference point: end-2017.

### Aggregate stress-test results and corporate vulnerability
- Debt-at-risk sensitivity:
  - Debt-at-risk is sensitive to both the profit and interest rate shocks.
  - The medium scenario assumes half the value of the shocks under the severe (adverse) scenario.
- Key aggregate outcomes under combined profit and interest rate shock (adverse scenario):
  - Share of vulnerable firms rises to 40 percent from 26 percent before the application of shocks.
  - Debt-at-risk rises to 38 percent from 23 percent before the application of shocks.
- Monetary conditions:
  - In both medium and adverse scenarios the corporate sector does not receive further gains from accommodative monetary conditions.
  - The combined profit and interest rate shocks reverse gains from recent economic recovery and monetary accommodation, taking corporate vulnerability indicators back to levels observed at previous crisis episodes.
- Consistency with banking stress tests:
  - These corporate sector results are consistent with the outcome of the banking sector solvency stress tests, where the majority of losses emanate from corporate credit risk.

### Drivers of corporate sector responsiveness (detailed stress test findings)
- The interplay of key factors identified in Appendix VI:
  - Severity of the adverse scenario:
    - The adverse scenario is more severe than recent crisis episodes with a cumulative nominal output loss of about 10 percent and sovereign yields remaining near historic highs for the duration of 3 years.
  - Monetary accommodation:
    - Monetary conditions remain accommodative as reflected in historically high ICRs, but the baseline scenario envisages some tightening in outer years while the adverse scenario does not allow for further accommodation.
  - Profitability:
    - Corporate profitability has been supported by cyclical factors and remains structurally below pre-crisis levels.
  - Concentration of debt:
    - Corporate debt is concentrated in low-profitability and low-ICR categories. In a low interest rate environment shocks can have unproportionately large effects.

### Conclusions on corporate sector resilience
- Overall assessment:
  - The Italian corporate sector remains vulnerable.
  - The non-financial corporate sector in Italy has been gradually recovering in the aftermath of the twin crises.
- Recent improvements and remaining weaknesses:
  - Inflows of corporate bad loans has declined sharply, remaining above the pre-crisis levels.
  - Accommodative monetary policy has contributed to nearly doubling the ICRs in four years.
  - Nevertheless, profitability is still below the pre-crisis average and net investment is slow to recover.
  - Corporate debt is concentrated near low-ICR and low-profitability segments and thus vulnerable to shocks.
- Impact of stress:
  - The stressed environment erodes fully the gains from monetary easing, lowering the ICR close to historical lows.
  - In response, the share of vulnerable firms as well as debt-at-risk increase to their earlier crisis-driven peaks.
- Implication:
  - The corporate sector remains an important source of vulnerabilities that can exacerbate asset quality concerns for the banking sector.

### Interconnectedness analysis — overview
- Scope:
  - Examined domestic interbank and inter-sectoral financial linkages as well as cross-border spillovers.
  - Financial interlinkages across sectors based on quarterly flow-of-funds data published by the Bank of Italy.
- Methods for cross-border spillovers:
  - Balance-sheet based analysis applying Espinosa-Vega and Solé (2010) methodology to BIS Consolidated Banking Statistics on cross-border bank exposures.
  - Market-data based contagion stress tests using daily equity returns of stock price indices (total market and banking sector indices) and sovereign CDS, following Diebold and Yilmaz (2014).
- Key high-level finding:
  - Direct domestic interbank contagion is limited owing to small interbank exposures, but exposures of Italy’s financial sector to the government and the nonfinancial corporate sector remain a key source of risk.

### Inter-sector and cross-sector contagion drivers
- Flow-of-funds analysis indicates cross-sectoral contagion risk due to:
  - Growing cross-sectoral exposures.
  - The significant role of the government bond market for banks, nonbanks (particularly insurance firms), and foreign investors.
  - Indirect links of households to sovereign risk intermediated through the financial system.
  - Important direct links of banks, households and nonresidents to corporate debt.
- Confidence effects:
  - Contagion risk within the banking sector could be triggered by confidence effects, though this effect is not considered in the quantitative analysis.

### Cross-border linkages and recent evolution
- Cross-border linkages:
  - Italy’s banking system embraces significant inward and outward cross-border spillovers, particularly with respect to other European countries.
- Shift in shock transmission:
  - Italian stock markets, bank returns, and sovereign CDS spreads have historically been net shock transmitters, but since May 2018 their importance declined and they became net shock receivers.
  - This suggests shocks originating externally are becoming increasingly relevant to the Italian financial system.
- Portfolio flows:
  - Foreigners have been selling Italian securities, while domestic residents have continued to build their net foreign asset position, exposing Italy relatively more to external shocks.
  - Note: Portfolio outflows intensified in 2018:Q2-Q3 totaling €72 billion, about 80 percent of which were related to government securities.

### Inter-sector financial linkages — key statistics and structure (flow-of-funds based)
- Data coverage: quarterly flow-of-funds data published by the Bank of Italy (via Haver) covering 1995:Q1–2018:Q1.
- Periods of focus:
  - 2008:Q1–2012:Q2 (pre-GFC to post-GFC, before OMT).
  - 2012:Q2–2018:Q1 (post-OMT to latest available).
- Institutional sectors analyzed (eight):
  - Non-financial: Government (GOV), Non-financial Corporations (NFC), Households and Non-profit Institutions Serving Households (HH), Rest of the World (RoW).
  - Financial: Bank of Italy (BdI), Other Deposit-taking Institutions (MFI), Insurers and Pension Funds (INP), Other Financial Institutions (OFI).
- Financial instruments:
  - 13 broad financial instrument categories aggregating 29 financial instruments.
- Inter-sectoral exposure levels:
  - Overall inter-sectoral balance sheet exposures stood at approximately 905 percent of GDP in 2018:Q1, up from 665 percent in 2008:Q1.
  - Sum of total assets (including intra-sectoral exposures) stood at 1020 percent of GDP in 2018:Q1, up from 775 percent in 2008:Q1.
- Composition and changes:
  - In 2018:Q1, currency and deposits were nearly a quarter of total issued instruments.
  - Bonds were the second largest component; central and local government bonds accounted for 60 percent of total bonds issued.
  - Equity instruments, mainly issued by NFCs, were nearly 17 percent of total issued instruments.

### Sectoral balance-sheet roles and exposures
- Government and corporates:
  - Government is the largest net borrower, followed by corporates.
  - Government’s net borrowing increased from 87 percent of GDP in 2008:Q1 to 135 percent in 2018:Q1.
  - NFCs’ liabilities to counterparties remained at 125 percent of GDP in 2018:Q1, as in 2008:Q1, although they have declined in absolute terms.
- Households:
  - Households have the largest net financial assets and shifted portfolios away from deposits and government bonds toward corporate equities, insurance and pension funds, and investments abroad.
  - Households’ direct exposure to the government is low at 5.5 percent of households’ assets, but they are indirectly exposed to sovereign risk via MFIs and INP.
- Pension funds and insurance companies:
  - At 2018:Q1, around 87 percent of pension funds’ and insurance companies’ funding came from households.
  - More than 45 percent of their assets were invested abroad in foreign currency bonds and equities, and an additional 34 percent in domestic government securities, increasing insurance sector exposure to sovereign risk.
- Banks:
  - Banks’ balance sheets expanded during 2008–12 and shrunk since 2012.
  - Households have large financial claims on the banking sector relative to their balance sheets (close to 25 percent).
  - Financial obligations of NFCs to banks continue to be significant, constituting one quarter of banks’ total assets.
- NFC funding structure and disintermediation:
  - NFCs’ net borrowing position remained broadly unchanged.
  - Significant disintermediation out of the bank lending channel occurred during 2012–18.
  - Corporate liabilities rebounded with more prominent funding from abroad and direct financing by households (largely in the form of shares and equities).

### Cross-border integration statistics and trends
- Cross-border integration levels:
  - Italy’s foreign assets increased from 107 percent of GDP in 2008:Q1 to 162 percent of GDP in 2018:Q1.
  - Foreign liabilities increased from 130 percent to 171 percent of GDP in 2018:Q1.
  - Net external financing declined from 22 percent of GDP in 2008 to close to 10 percent of GDP in 2018.
- Composition changes since 2012:
  - Increase in foreign assets driven more by foreign investment of Italian nonbank institutions (INP, NFCs, HH) and less by Italian banks.
  - Increase in foreign liabilities driven by increased foreign participation in Italian government bonds since 2012:Q2 and in NFCs.
- Role of Rest of the World (RoW):
  - RoW is an important source of government funding: RoW investment in Italian sovereign bonds increased from 40 percent to 53 percent of GDP in 2018:Q1.
  - External financing accounts for 31 percent of the government’s total funding.

### Intra-corporate exposures and contagion channels
- NFC intra-sector exposures:
  - At 2018:Q1, close to 43 percent of total intra-corporate exposures is composed of shares and equity, with the remainder as trade credit and loans.
  - Large intra-corporate balance sheet exposures (47 percent of GDP in 2018:Q1) may represent a contagion channel if corporates fail to service reciprocal claims, potentially spilling over to the financial sector.

### Interbank network contagion assessment (methodology and assumptions)
- Network model: Espinosa-Vega and Solé (2010) applied to a matrix of bilateral interbank exposures.
- Default rule:
  - A bank is assumed to default when the bank’s CET1 ratio drops below 4.5 percent of risk-weighted assets (regulatory requirement).
- Contagion mechanics:
  - Default affects other banks through asset and liability exposures.
  - Asset-side: loss given default of 100 percent is applied.
  - Liability-side: banks are assumed forced to sell assets at a discount to replace lost funding, incurring additional loss.
  - Funding shock parameters:
    - Only 65 percent of the lost funding can be replaced.
    - Banks need to sell assets with a valuation loss of 50 percent to cover the lost funding.
- Caveat:
  - Contagion is modeled as instantaneous; the model assumes no time for banks to take mitigating measures between rounds of contagion.

*Source: IMF staff analysis and flow-of-funds data as presented in the provided content.*

### 153.      The matrix of direct interbank exposures contains the eleven SIs and sixty-two LSIs.

### 1itaea2020008 - 153.      The matrix of direct interbank exposures contains the eleven SIs and sixty-two LSIs.

### Interbank exposures: scope and composition
- The interbank matrix includes the eleven SIs and sixty-two LSIs and is based on supervisory data covering each bank’s interbank assets and liabilities vis-a-vis each of the other 72 banks as of end-June 2018.
- The interbank matrix includes interbank loans and securities; data on off-balance sheet exposures and interbank equity and capital participation was not available and therefore was not included.
- Interbank loans and securities exposures among the 72 banks amount to 1.05 percent of banking system assets and 1.6 percent of GDP.
- About half of interbank exposures consists of interbank loans, with the other half consisting of interbank securities.
- The majority of interbank assets and liabilities are vis-à-vis the SIs, comprising 89 percent and 78 percent of total interbank assets and liabilities, respectively.

### Aggregate positions and funding roles
- On an aggregate basis, LSIs are net lenders in the interbank market, while SIs are net borrowers.
- LSIs’ main source of funding is retail deposits (i.e., lower funding costs); LSIs allocate 85 percent of their total interbank assets to lending to SIs.
- LSIs also lend to other LSIs; LSIs hold securities issued by SIs.
- SIs are net lenders in the interbank loans market, lending mostly to other SIs.

### Network analysis (balance-sheet based) — contagion through direct interbank exposures
- The balance-sheet based network analysis reveals that contagion risks stemming from interbank exposures are very limited.
- Banks generally have sufficient capital to withstand interbank contagion through their direct balance-sheet exposures.
- Derivatives exposures and off-balance sheet interbank exposures were not included in the analysis; these excluded exposures are equivalent to 5.6 percent of total assets and 8.5 percent of GDP.
- Results are robust to changes in parameter values and different combinations of shocks (credit only, or credit plus funding shocks).
- Simulations consider only effects arising from individual bank defaults and do not measure systemic implications from simultaneous distress in many banks.

### Vulnerabilities by bank size and simulation results
- The smaller LSIs (with less than 1 percent asset share in total LSI assets) are most vulnerable to interbank spillovers due to relatively low capital against interbank exposures.
- SIs are the most systemically important; a hypothetical default of a SI could, on average, reduce the capital of other banks by 9.3 percent.
- LSIs would lose the highest share of initial capital (on average, across all simulations) due to interbank contagion.
- Simulation counts and contagion events:
  - In eleven simulations where each of the SIs defaulted, four SIs triggered the default of one LSI, and one SI triggered the default of 3 LSIs. (Footnote: One LSI defaulted in two simulations.)
  - In sixty-two simulations where each of the LSIs defaulted, one LSI triggered the default of another LSI.
- The index of vulnerability (percentage point of average capital loss due to the default of all other institutions) would be 5 percent at most.

### Cross-border interconnectedness: approaches and key findings
- Two complementary approaches were used:
  - The Espinosa-Vega and Solé (2010) methodology with BIS Consolidated Banking Statistics (CBS) to examine cross-border bank exposures and direct banking sector linkages.
  - The Diebold and Yilmaz (2014) methodology with daily equity returns of stock price indices (total market and banking sector indices) and sovereign CDS to examine interconnectedness between Italy and countries with strong financial and trade linkages with Italy.
- CBS data (2018:Q2) summary:
  - Total consolidated foreign claims of Italian banks stood at US$0.65 trillion in 2018:Q2.
  - 53 percent of the claims were against foreign nonbank private sector, 22 percent against the public sector, and the remainder against banks.
  - Italian banks’ foreign claims are concentrated in Germany, Austria, Spain, France, the United States and the United Kingdom.
  - Total foreign banks’ claims on Italy stood at about US$0.66 trillion in 2018:Q2; about 53 percent of these are against the Italian nonbank private sector and 30 percent against the public sector.
  - French banks’ claims on the Italian nonbank private sector, Italian banks, and public sector are close to 62 percent, 42 percent, 32 percent of total foreign claims in these sectors of Italy, respectively.

### Network analysis using BIS CBS — simulation results (Espinosa-Vega and Solé, 2010)
- Two sets of simulations using BIS-CBS for 2018:Q2:
  1. Simulation 1: reporting banks’ exposure to foreign banks only, considering both credit and funding shocks.
  2. Simulation 2: impact of credit shock to the total exposure of the banking sector, including claims to banks, government and the nonfinancial sector.
- Simulation 1 (bank exposure, credit and funding shocks):
  - The Italian banking sector is closely linked to other European banking systems through interbank exposures.
  - Credit and funding shocks in the Italian banking sector have the largest outward spillover to banks in Luxembourg, Turkey and Austria (measured by percentage of capital loss in a banking system due to the default of all bank-to-bank exposures).
  - Inward spillovers to Italy are largest from banks in France, Germany, the United States, the United Kingdom, and Spain.
- Simulation 2 (total exposures, credit shock):
  - Credit shocks originating from France, Germany, Spain and the United States would have the largest impact on Italian banks.
  - Inward spillover to Italian banks is higher when all exposures are considered compared to bank exposure only.
- Coverage note: The simulation sample consists of 21 countries: Austria, Australia, Belgium, Canada, Switzerland, Germany, Spain, Finland, France, the United Kingdom, Greece, Ireland, India, Italy, Japan, Luxembourg, The Kingdom of the Netherlands-Netherlands, Portugal, Sweden, Turkey and the United States. Data availability for Luxembourg and Turkey is relatively low.

### Contagion analysis using market data (Diebold and Yilmaz, 2014)
- Methodology and data:
  - The Diebold and Yilmaz (2014) approach decomposes the variation in an asset’s volatility into contributions from shocks to other assets; financial market spillover from country A to B is the share of variation in B’s equity return shocks attributable to shocks in A’s equity returns.
  - The analysis uses daily market information on stock (total market and banking sector) and sovereign CDS covering 19 countries.
  - Data source: MSCI and DataStream indices from July 2005 to October 2018 at daily frequency.
  - To control for differences in trading hours, equity returns are computed as the average two-day log returns for equity prices.
- Caveats:
  - The methodology cannot identify causality or the source(s) of spillovers.
  - The methodology does not distinguish the direction of co-movement in equity prices; negative and positive equity price responses are measured in the same way.
  - The approach complements balance-sheet analysis; convergence between the two analyses adds reassurance.
- Empirical sample for market-data analysis: Australia, Brazil, Chile, China, France, Germany, India, Italy, Japan, Mexico, The Kingdom of the Netherlands-Netherlands, Portugal, Spain, Switzerland, Sweden, Turkey, Russia, the United Kingdom and the United States.

*Source: IMF staff calculations and supervisory data as presented in the cited content unit.*

### 173.      Italian banks are highly connected with European banks from Spain, France, Germany,

### 1itaea2020008 - 173.      Italian banks are highly connected with European banks from Spain, France, Germany,

### Cross-border connectedness (banking, equity, sovereign CDS)
- Italian banks are highly connected with European banks from Spain, France, Germany, the United Kingdom, and Portugal.  
- Both inward and outward spillovers between Italy and these countries is significant.  
- Strong linkages are attributed to balance sheet linkages (BIS exposure data) and macroeconomic, monetary policy, and financial regulations interlinkages.  
- Banking indices from Spain, France, and the United Kingdom appear to be the source of return connectedness for Italian banks; Portugal, Germany and Sweden are found to be the recipients.  
- Spillover matrices do not differ significantly when applied to the total stock market index or to the bank stock indices, implying limited differentiation of international spillovers across sectors and implicit intersectoral contagion.

### Net contributions to systemic risk (returns and sovereign CDS)
- Italian banks are an important source of net return connectedness in the global banking system.  
  - Net directional connectedness of equity returns is measured as: outward spillover from the country’s banking index to the global banking system minus the inward spillover from the system to the country.  
  - Together with France, Spain and the United Kingdom, Italian banks are found to be important net contributors to return connectedness in the global banking system.  
- Bank equity indices in the United States, Sweden, Brazil and Germany also appear to be important sources of return connectedness.  
- Banks in Asia and Latin America (other than Brazil) tend to be recipients of return connectedness.  
- Italy’s net outward spillovers to total stock market prices is significant but relatively lesser than across the global banking system.  
  - France, the United States, Germany, the United Kingdom and the Kingdom of the Netherlands-Netherlands are the largest net contributors to return connectedness in global stock markets, followed by Sweden and then Italy.  
  - Italy’s financial markets are important net contributors to stock markets in Asia.  
- Italy’s sovereign credit risk is a major source of spillover to global CDS markets.  
  - Italian sovereign CDS is the second largest net contributor to sovereign CDS markets, only after Mexico.  
  - Italian sovereign CDS appears to be the source of net spillovers to all countries in the sample and shows very tight inward and outward spillover connections to neighbors, particularly to Spain and Portugal in CDS markets.  
  - There is strong regional clustering; inward and outward sovereign credit risk spillovers from Italy are likely to spread mainly within the EA.

- Empirical basis noted: Results are based on the Diebold and Yilmaz (2014) approach using daily (log) returns in stock indices (total and banking) and sovereign CDS from July 2005 to October 2018.

### Dynamics in net return connectedness (time variation)
- Net outward spillovers from Italy have been on a decline since early 2018.  
- Historically, Italian stock markets, banks and sovereign CDS have been net shock transmitters to stress in the financial system.  
- Since May 2018, returns on Italian stocks and sovereign CDS have increasingly become net shock receivers rather than net shock originators, indicating externally originating shocks (particularly from Spain, the United Kingdom, France and Germany) are becoming stronger and more relevant to the Italian financial system.  
- Net outward spillovers from the Italian banking sector remained positive during the Italian election period in the first half of 2018 and have started to pick up in other markets since September 2018.

### Risk Assessment Matrix — selected risks and impacts
- Sharp rise in risk premia.  
  - Overall Level of Concern: High. Relative Likelihood: High.  
  - Expected impacts: Significantly higher borrowing costs for the sovereign leading to devaluation of sovereign bonds and bank losses; commensurate rise in banks funding, further squeezing profitability and leading to forced deleveraging or further underprovisioning.  
- Weaker-than-expected global growth.  
  - Overall Level of Concern: High. Relative Likelihood: Medium.  
  - Expected impacts: Slower output growth, slow-down in potential output; lower growth weakening public debt sustainability; weaker investment and persistent long-term employment further damaging private balance sheets, leading to further formation of NPLs.  
- Increase in Italy-specific risk premia and widening of spreads relative to other Euro-area sovereigns due to political uncertainty.  
  - Overall Level of Concern: Medium. Relative Likelihood: High.  
  - Expected impacts: Higher borrowing costs for the government with little ECB support; large losses on bank balance sheets due to sovereign exposures leading to deleveraging; higher borrowing costs for the private sector as banks shift costs to customers.

### Stress Testing Matrix (STeM) — Banking sector: Solvency Stress Test (overview)
- Institutional perimeter and coverage:
  - Exercise: Top-Down by FSAP team.  
  - Institutions included: All SIs (9) in the sample and select LSIs (62 institutions). The latter are only subject to sensitivity analysis.  
  - Market share: Total coverage is around 76 percent, with 68 percent for SIs and 8 percent for LSIs.  
  - Data and baseline date: Latest data: December 2018 for SIs and June 2018 for LSIs. Supervisory data include balance sheet information, COREP and FINREP, Short Term Exercise (STE), Expected Default Frequency from Moody’s Analytics and large exposure (LE) templates provided by the authorities, plus other supervisory information (NPLs by portfolio, PDs by portfolio from credit register, PDs for consumer loans from CRIF, funding and lending rates by asset and funding portfolios). Transparency templates for banks in the 2019 EBA stress test sample and equivalent data for other banks were also provided. Scope of consolidation: banking activities of the consolidated banking group for banks headquartered in Italy. Coverage of sovereign and non-sovereign securities exposures: fair value accounts (FVTPL and FVTOCI) and valued respectively at amortized cost (AC).
- Methodology and channels:
  - FSAP team satellite models and methodologies; balance-sheet regulatory approach.  
  - Baseline and adverse scenario use fully load IFRS 9 capital ratios (i.e., as if transitional arrangement for IFRS 9 had not been applied).  
  - Market risk treated as an add-on component with separate calibration; impacts on capital resources (via profit and loss or OCI) and on capital requirements (RWA).  
  - Losses for securities based on duration approach. For equities and associated hedges, the floor set by the EBA methodology: ∆Eq=1.5*(-0.20%(Eq^Long+Eq^Short)).  
  - For IRB portfolios, provisions are calculated for mortgage, financial institution, corporate large firms, corporate SMEs, and consumer loans. Expected losses = projected point-in-time PDs × LGDs × projected EADs. Point-in-time PDs are projected using regression models with macro variables; PiT PDs used for provisioning, TTC PDs used to project RWAs. PiT LGD based on historical recovery rates of NPLs. The same PiT LGD is used for baseline and adverse scenario.  
  - For credit risk on fixed-income instruments differentiate between AC and fair value holdings; credit risk for fair value holdings embedded in market risk methodology; for AC securities, provision calculated for any increase in credit risk.
- Stress test horizon:
  - 2018:Q4–2021:Q4 (3 years)
- Tail shocks and scenarios:
  - Baseline scenario based on the April 2019 WEO macroeconomic projections.  
  - Adverse scenario designed using staff’s Global Macrofinancial Model (GFM) to capture key risks in the RAM.  
  - Sensitivity analysis: single-factor tests including top 3 to 5 exposures failing for SIs; three-factor shocks for LSIs (interest rate risks (IRRBB), decrease in sovereign bond prices, credit losses from asset quality deterioration). Sensitivity on increase in RWA of SME loans (SIs only).
- Risks covered and behavioral assumptions:
  - Risks covered: credit, market (equity, exchange rate and interest rate risks), sovereign (repricing and spread risks), funding risks, and interest rate risk on the banking book. Concentration risks covered via sensitivity analysis.  
  - Behavioral adjustment: quasi-static balance sheet growth in line with nominal GDP path; floor of zero percent on balance sheet growth to prevent deleveraging (binding in adverse scenario). For RWAs, standardized and IRB portfolios differentiated; interest income from non-performing loans is not accrued. Banks assumed not to issue new shares or repurchases during horizon; dividends assumed to be paid out at 30 percent of current period net income after taxes (only if net income is positive) by banks in compliance with supervisory capital requirements.
- Regulatory and calibration parameters:
  - Calibration based on IMF staff credit models and satellite models projecting PiT PDs from credit register series (consumer loan PDs from CRIF). For exposures outside Italy, Moody’s EDF data used. TTC PDs estimated as average PDs over 8 quarters.  
  - Regulatory minima: CET1 (4.5 percent) and include requirements due to systemic buffers for O-SII. Evaluated banks’ total capital adequacy ratio against the 8 percent level, Tier 1 capital ratio against the 6 percent benchmark and leverage ratio against the 3 percent Basel III minimum requirement. The same hurdle rate used for baseline and adverse scenario; hurdle rates do not include capital conservation, countercyclical buffers or pillar 2 requirement. For LSIs sensitivity test, established hurdle rate is 7 percent CET1 ratio. Banks ending the horizon below relevant hurdle rates are considered to have failed the test.
- Reporting outputs:
  - Reported outputs include evolution of capital ratios for the system, for LSIs by regional groups; impacts by drivers (profit components, losses), capital shortfall in euros and percent of nominal annual GDP, number of banks and percentage of assets below regulatory minimum or minimum leverage ratio.

### Stress Testing Matrix — Banking sector: Liquidity Stress Test (overview)
- Institutional perimeter and coverage:
  - Exercise: Top-Down by FSAP team.  
  - Institutions included: All SIs (11) and select LSIs (62 institutions) (LSIs only in sensitivity analysis).  
  - Market share: Total coverage is around 82 percent, with 74.1 percent for SIs and 8 percent for LSIs.  
  - Data and baseline date: Latest data: October 2018 for SIs and end-June 2018 for LSIs. Source: supervisory data (LCR, NSFR and ALMM Maturity Ladder template). Scope of consolidation: banking activities of the consolidated banking group for banks headquartered in Italy.
- Methodology and channels:
  - Basel III LCR and NSFR, cash-flow based liquidity stress test using maturity buckets by banks, incorporating contractual and behavioral (where available) flows with assumptions about combined interaction of funding and market liquidity and varying levels of central bank support. Liquidity test in total currency and for large depositor withdrawals.
- Risks and buffers:
  - Risks: Funding liquidity; market liquidity; counterparty/depositor concentration risk (withdrawal of top 1, 3 and 5 depositors).  
  - Buffers: Counterbalancing capacity including market and central bank facilities; expected cash inflows included in cash-flow and LCR analyses.
- Tail shocks and calibration (size of the shock):
  - Run-off rates calibrated to reflect system-wide deposit runs and dry-up of unsecured wholesale and retail funding, with additional run-off for non-resident deposits; calibration based on historical events, recent international experience and IMF judgment.  
  - Retail scenario assumptions: (i) 10 percent run-off rates for stable retail deposits and 15–20 percent for less stable retail; (ii) 75 percent (40 percent) run-off rates for non-operational (operational) deposits not covered by Deposit Guarantee Scheme; and (iii) 10 percent haircut on government securities for HQLA calculation.  
  - Wholesale scenario assumptions: (i) 100 percent run-off rates for wholesale funding from other financial institutions; (ii) 50 percent run-off rates for operational deposits not covered by Deposit Guarantee Scheme; (iii) 50 percent run-off by non-operational deposits not covered by Deposit Guarantee Scheme; (iv) 10 percent haircut on government securities for HQLA calculation.  
  - Liquidity shocks simulated for 1-month for both LCR and cash-flow approaches, and 5-day and 3-months for cash-flow based approach.  
  - HQLA haircuts calibrated consistent with market shock for investment securities and money market instruments in solvency stress test.

*Source: IMF staff calculations.*

### 5. Regulatory and

### 5. Regulatory and Market-Based Standards and Parameters

### Regulatory standards
- Consistent with Basel III regulatory framework (LCR and NSFR).
- Liquidity shortfall by bank.

### Reporting Format for Results — Liquidity
- Liquidity ratio or shortfall by groups of banks and aggregated (system wide).
- Number of banks that still can meet or fail their obligations.

### Bank and non-bank sector: Contagion Analysis

### 1. Institutional Perimeter
- Institutions included:
  - All SIs (11) and select LSIs (62 institutions).
- Market share:
  - Total coverage is around 82 percent, with 74.1 percent for SIs and 8 percent for LSIs.
- Data and baseline date:
  - Supervisory as of end-June, 2018; and market data.
  - BIS consolidated banking statistics.
  - Flow-of-funds quarterly data (2008:Q1-2018:Q1), for analysis of cross-sectoral linkages.

### 2. Channels of Risk Propagation
- Methodology:
  - Balance-sheet model: Interbank and cross-border network model by Espinosa-Vega and Solé (2010).
  - Market-based model: Diebold-Yilmaz (2014) generalized forecast variance decomposition approach.

### 3. Tail shocks
- Size of the shock:
  - Pure contagion: hypothetical default of institutions.
  - Default threshold: banks would default if their CET1 capital ratios fall below 4.5 percent (regulatory minimum).

### 4. Reporting Format for Results — Contagion Outputs
- Number of undercapitalized institutions in distress;
- Capital shortfall systemwide, by bank and by group: contagion and vulnerability scores;
- Amplification and cascade effects, direction and size of spillovers within the network.
- Net spillovers due to interconnectivity (market-based).
- Market-based analysis: Varian Decomposition (spillover contribution to equity prices).

*Italic source attribution: Chapter 5 and related appendices from the provided IMF content.*

### Appendix III. PD Estimation Models

### 1. PD proxy, model form, and data
- Proxy used: ratio of new non-performing loans to total performing loans as proxy for probability of default (PD).
- Separate satellite models for: corporate large firms, corporate SME, mortgage, consumer loans and financial institution.
- Point-in-time (PiT) PDs projected using regression models with macro variables as independent variables.
- Data: provided by Banca d’Italia on a quarterly and aggregate basis, from 2006:Q1 to 2018:Q4.
- Conditional PD forecasts attached to starting point of each individual bank’s PDs to generate full path under baseline and adverse scenarios.
- Logit transformation applied to original PD:
  - Y_i,t = log( PD_i,t / (1 − PD_i,t) )
- Estimated model (logit-transformed PDs as linear function):
  - Y_i,t = α + β Y_{i,t−k} + δ X_{t−S} + ε_{i,t}
- Projected logit PDs transformed back to PD space.

### 2. Estimation framework
- Linear Bayesian Model Averaging (BMA) used to remove model uncertainty and average over best models according to approximate posterior model probability.
- BMA enables sign restriction and robustness of out-of-sample forecasts conditional on sample size.

### 3. Model selection criteria
- Five information criteria used: R-square, Durbin Watson statistics, number of significant variables with high posterior inclusion probability, quality of in-sample forecast, size of the impact in the forecasting horizon.
- Ideal candidate: relatively high R-square, DW statistics between 1.5 to 2.5, small root-mean-square-error and historically coherent size of impact under stress.

### 4. Key macro drivers and estimation diagnostics
- Relevant variables: real domestic output, unemployment rate, inflation, short-term and long-term interest rate.
- Table AIII.1 (selected diagnostics by segment) — preserved values:
  - Number of observations: 52, 52, 52, 52, 60 (by column)
  - Number of lags of independent variables: 2, 2, 2, 2, 2
  - R square: 0.91, 0.56, 0.91, 0.91, 0.99
  - Durbin Watson Statistics: 1.78, 1.91, 2.35, 2.11, 2.23
- Selected coefficient signs and significance marks (as presented):
  - GDP growth, percent, yoy: -0.01, -0.52*, -0.52*, -0.01, -1.03*
  - Unemployment rate, percentage point change, yoy: 0.29*, 0.01, 0.63*, 0.99*, 0.02
  - Inflation rate, percent, yoy: 0.00, -0.17*, 0.00, 0.00, 0.00
  - Long term sovereign bond yield, percent: 0.07, 0.15*, 0.07, 0.01, 0.00
- Note: Table headings include Mortgage, Financial Institution, Large Firms, SME, Consumer Loans.

### Appendix IV. Funding and Lending Rate Estimation

### 1. Data and scope
- Bank funding costs estimated based on funding rate for new deposits (front book) for each bank.
- Data: authorities provided quarterly from 2006:Q1 to 2018:Q4.
- Macroeconomic data sources: IMF IFS, IMF WEO, Bloomberg, Haver Analytics.
- Adverse scenario macro series followed scenario set for this stress test.

### 2. Method and structure
- Bank interest rates on new business estimated and used for interest risk assessment on the banking book.
- BMA methodology used for satellite models to estimate aggregate funding and lending rates on portfolio level (retail and wholesale deposits, debt securities, household and corporate loans).
- Bank-specific interest rate paths constructed by attaching period changes of aggregate rates to bank-specific starting points.

### 3. Explanatory variables and cross-country factors
- Inputs similar to credit risk models: GDP growth, inflation, exchange rate, unemployment, short- and long-term interest rates.
- Explanatory variables include Italy-specific, EA- and U.S.-based indicators to account for domestic and foreign exposures.

### 4. Interest rate passthrough and sequencing
- Funding cost included as explanatory factor in projection of lending rate to allow partial passthrough.
- Model performed sequentially: first estimate funding rate, then use as input for lending rate projection.

### 5. Inclusion of PDs
- PDs mapped into respective lending rate categories to capture banks’ higher interest charge on borrowers with high default risk.

### 6. Choice of forecast percentile and results on NIM
- To be conservative, selected lower bound within forecast confidence band:
  - 5th percentile for the adverse scenario
  - 25th percentile for the baseline
- Resulting average decline on the net interest margin:
  - 0.07 percent under the baseline
  - 0.18 percent under the adverse scenario

### 7. Main contributing variables and long-run multipliers
- Money market rate and long-term sovereign yield are main contributors in projections of bank interest income and funding costs.
- On funding side: 3-month money market rate and 10-year domestic sovereign bond yield spread explain majority of movement in interest expense.
- On lending side: U.S. long term sovereign bond yield and funding cost associated with retail term deposits also significant.
- Long run passthrough sizable from Italy sovereign bond yield and short-term interest rate to funding rates, particularly:
  - Government bond yield to retail term deposit
  - Short-term rate to wholesale term deposits, retail and wholesale overnight deposits

### Key diagnostics from Table AIV.1 (selected preserved values)
- Number of observations (per column): 52 for each portfolio shown.
- Number of lags of independent variables: 2 (per column).
- R square (per column): 0.98, 0.97, 0.96, 1.00, 1.00, 0.96, 1.00, 0.99
- Durbin Watson Statistics (per column): 1.82, 2.32, 1.74, 1.64, 1.78, 1.93, 1.79, 2.24
- Selected coefficient signs and significance marks (as presented):
  - Italy sovereign long term yield spread against German bund, percent: 0.9*, 0.16*, 0.08*, 0.6*, 0.43*, 0.000.09*, 0.09*
  - Euro Area short term interest rate, percent: 0.7*, 0.98*, 0.44*, 1.18*, 1.18*, 0.03, 0.54*, 0.46*
  - Interest rate, retail term deposit, percent: 0.67*, 0.6*, 0.63*
  - U.S. Long term sovereign bond yield, percent: 0.000.000.000.000.000.43*, 0.11*, 0.02

### Appendix V. Characteristics of Effective Interest Rates in Micro-Data
- Table AV.1 key reported coefficients and stats (preserved values as presented):
  - Lending rate on new NFC loans: 0.735*, 0.081, 0.432, 0.493***, 0.299***, 0.291***, 0.925***, 0.844***, 0.840***
  - Lagged dependent variable (t-1): 0.191*, 0.190*, 0.058***, 0.058***, 0.044***, 0.044***
  - Constant examples: 0.186***, 0.157***, 0.386, 0.012***, 0.016***, 0.021, 0.028***, 0.027***, 0.020
  - Observations (various columns): 625, 2407, 46240, 134616, 826100, 160448, 132506, 810398, 496130, 9377, 217587, 700734
- Note: standard errors in parentheses, ***, **, and * denote significance at 1, 5, and 10 percent, respectively. Columns 3, 6, and 9 include full set of fixed effects for firm size, sector, and geographic area.

### Appendix VI. Non-Financial Corporate Stress Test Results
- Table AVI.1 selected preserved entries:
  - Baseline:
    - ICR < 1: 3.326
    - ICR < 2: 3947
    - ICR < 3: 234047
  - Adverse scenario:
    - Profit shock: 2.7324452254249
    - Interest shock: 1.8365261365262
    - Combined shock: 1.5405563385463
  - Medium scenario:
    - Profit shock: 3.2284149244148
    - Interest shock: 2.4304655314756
    - Combined shock: 2.2324856324756
- Table labels: Share of Firms, Share of Debt, Median ICR (as presented in source).

### Appendix VII. Cross-Sectoral Interconnectedness

### Grouping of sectors (Table AVII.1 preserved coding)
- Nonfinancial Corporations (S.11) — NFC
- Monetary Financial Institutions Excluding Central Bank (S.122+123) — MFI
- Bank of Italy (S.121) — BdI
- Other Financial Intermediaries (S.125) — OFI (excluding non-MMF Investment)
- Non-MMF Investment Funds (S.124)
- Captive financial institutions & money lenders (S.127)
- Financial Auxiliaries (S.126)
- Insurance Companies (S.128) — INSP
- Pension Funds (S.129)
- Central Government (S.1311) and Local Government (S.1313) — GOV (S.13 – General Government)
- Social Security Funds (S.1314)
- Households (S.14) and Nonprofit Institutions Serving Households (S.15) — HH
- Rest of the World (S.2S) — RoW

### Cross-sectoral network presentation
- Data presented as cross-sectoral networks by financial instrument and on a non-consolidated basis to highlight intra-sector exposures.
- Cross-sectoral exposures presented as adjacency matrices: holding sectors in rows and issuing sectors in columns.
- Non-consolidated data used to identify and analyze relevant intra-sectoral activity for financial stability monitoring.
- Methodology references: steps and assumptions to infer sectoral exposures from Italian flow-of-funds data per Garcia-Macia (2018).

### Financial Network Matrix
- Table AVII.2: Italy: Financial Network Matrix (Gross Bilateral Exposures, as of 2018:Q1) — data source: Banca d’Italia (via Haver) and IMF staff calculations.

### Appendix VIII. BIS International Banking Statistics

### 1. BIS statistics types
- BIS collects two types: Consolidated Banking Statistics (CBS) and Locational Banking Statistics (LBS).

### 2. Locational Banking Statistics (LBS)
- LBS focus on all banking offices (domestic- and foreign-controlled) resident in the reporting country and cover unconsolidated cross-border and local positions of these banks.
- Data based on either residence or nationality of reporting institution (Italian resident banks include both Italian-owned banks and foreign-owned branches and subsidiaries resident in Italy).
- LBS correspond to compilation of national accounts, balance of payments, and external debt statistics.

### Figure
- Figure VIII.1: Italy: Structure of BIS Consolidated Banking Statistics (source: Bank for International Settlements (BIS)).

*Italic source attribution: Chapter 5 and appendices (III–VIII) from the provided IMF PDF content.*

### 3.      The CBS focus on banks headquartered in the reporting country and cover banks’

### 1itaea2020008 - 3.      The CBS focus on banks headquartered in the reporting country and cover banks’

### Consolidated Banking Statistics (CBS): scope and measurement
- CBS focus on banks headquartered in the reporting country and cover banks’ consolidated on-balance sheet claims and selected off-balance sheet exposures to counterparty countries and sectors.
- Positions are reported by head offices in their home country and include all branches and subsidiaries on a worldwide consolidated basis, net of inter-office accounts.
- The dataset is available on:
  - an immediate borrower basis (CBS-IBB) based on the country of the first counterparty exposure, and
  - an ultimate risk basis (CBS-URB) based on the country where the final risk resides after taking into account risk transfers.
- Three different types of claims are included in the CBS: C, D, and E (as listed in the source text table).
- Classification of total foreign claims:
  - Under CBS-IBB: divided into international claims and local claims in local currency.
  - Under CBS-URB: categorized into cross-border claims and local claims.
- Use case selection:
  - The LBS are used to analyze capital flows between countries.
  - The CBS are used to measure the country risk exposures of internationally active banking groups.
  - This note focuses on “risks” that banks transfer in and out of the Italian economy, rather than their role as a conduit; therefore the CBS are used instead of the LBS.

### Market-based spillover analysis: data and VAR methodology
- Methodological basis:
  - The connectedness methodology draws from Diebold and Yilmaz (2014) for market-data analyses.
  - Measurement starts with estimating a Vector Auto Regression (VAR) with specification Y = A(L)Y + ε, where Y is a vector of daily range-based volatility in equity returns, A(L) is the lag polynomial, and ε is an error term.
- Pairwise bilateral linkages:
  - Calculated from the fitted VAR model using the Generalized Variance Decomposition (GVD), based on Pesaran and Shin (1988).
  - Spillover measures are the percent contribution of entity A to the H-step-ahead forecast error variance of entity B (entities are banks and NBFIs in this analysis).
  - Advantage: does not require any assumption on the order of the variables (unlike Cholesky ordering).
  - The horizon H captures dynamic connectedness (analogous to contagion) as opposed to purely contemporaneous connectedness.
- Network spillover measures and analogies:
  - From-degree measure: captures exposures of individual firms to shocks from the network (inward spillover), analogous to Marginal Expected Shortfalls (MES).
  - To-degree measure: captures contributions of individual firms to network events (outward spillover), analogous to Delta CoVaR.
  - Net-degree measure: difference between to- and from- measures; describes the relative contribution to financial stress from each financial firm.
- Estimation specifics and robustness:
  - Results reported are based on generalized impulse response functions.
  - VARs are estimated with three lags.
  - Connectedness is calculated for a 10-day horizon.
  - Spillover measures are found to be robust to the choices of lag, forecast horizon, as well as to equity returns obtained from H-shares and A-shares of dual-listed FIs.
  - The VAR model turns to LASSO to estimate the long-run correlation network.
  - Note: Diebold and Yilmaz (2009) highlight that volatility connectedness measures display spikes along with trend, unlike returns connectedness which tends to trend smoothly.

### Sample composition and VAR specifications (Table IX.1)
- Market Source Period Sample:
  - Stock index (Total market): MSCI 07/12/2005-10/31/2018 19
  - Stock index (Bank index): Datastream 07/12/2005-10/31/2018 18
  - Sovereign CDS: Datastream 01/05/2009-10/31/2018 16
- Additional sample and estimation notes:
  - For the dynamic version, the rolling sample windows size is set to 150 for equity indices and bank stock returns and 200 for sovereign CDS and FX series.
  - Number of lags in the VAR is set to 3 and the forecast horizon is 10 for all series.
  - The sample includes all countries for total market stock indices.
  - For banking indices, Netherlands, Switzerland and Russia are excluded due to data constraints.
  - Australia, India and Switzerland are excluded from the CDS exercise due to similar constraints.

*Italic: Source — Excerpt from 1itaea2020008 (IMF PDF content provided).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2020/english/1itaea2020008.pdf_