## Georgia FSAP: Banking System Stress Test (Content unit 1geoea2021005)

## Source details

**Canonical URL:** [Georgia FSAP: Banking System Stress Test (Content unit 1geoea2021005)](https://www.imf.org/-/media/files/publications/cr/2021/english/1geoea2021005.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2021/english/1geoea2021005.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2021/english/1geoea2021005.pdf.json)

---

### Executive summary — background and system structure
- Total bank assets are about 115 percent of GDP.
- The three largest banks account for about 85 percent of total assets.
- 14 out of the 15 banks have some foreign ownership; there are no public banks.
- About half of bank loans and two-thirds of bank liabilities are denominated in FX.
- Non-resident deposits account for about one-fourth of retail deposits and 7½ percent of banks’ balance sheets.
- International reserves stand at adequate levels according to standard metrics, but persistent current account deficits have led to the accumulation of substantial foreign liabilities.
- As of end-December 2020:
  - Retail deposits represent about 36 percent of total liabilities.
  - Deposits from non-financial companies add an additional 32 percent.
  - The share of FX deposits in total deposits has remained slightly above 60 percent.
- The system displays a structural shortage of lari liquidity; banks partially roll over short-term central bank funding.
- COVID-19 macroshock:
  - Economy contracted by 6.1 percent in 2020.
  - Nominal FX depreciated by 17.7 percent vis-à-vis the USD in March 2020; the year closed with a more moderate average nominal depreciation of about 10 percent.
  - Unemployment rate increased to about 18½ percent.
  - Fiscal deficit ballooned to 8.9 percent of GDP and public debt jumped from about 40 percent of GDP in 2019 to 60 percent of GDP in 2020.

### Stress test scope, methodology, and focus
- Exercises:
  - Top-down stress tests carried out jointly by the NBG and the FSAP teams during January‒April 2021.
  - Bottom-up exercises were not implemented at the authorities’ request due to COVID-19 operational challenges.
- Primary focus areas:
  - Credit and liquidity risks, with emphasis on FX-induced credit risk and concentration risk from lending to large corporates or sectoral clusters.
  - Liquidity assessment included currency-specific LCR analysis and potential sudden calls on liabilities, including by non-residents.
- Scenarios:
  - Baseline and two distressed scenarios: Extended Pandemic, and Capital Outflows (Extended Pandemic + adverse external financial environment).
  - Sensitivity analyses on market, interest rate, and FX shocks; liquidity shocks calibrated using dynamics at the peak of the 2008‒09 global financial crisis.
- Hurdle rates and key parameters:
  - Basel CAR: 8 percent; T1: 6 percent; CT1: 4.5 percent.
  - Overall LCR requirement: 100 percent; currency-specific LCR: 100 percent for FX, 75 percent for lari.
  - Semi-static balance sheets; dividend payout ratio assumed to be 50 percent in stress tests.
  - Cut-off date: December 2020.

### Key results — credit quality, losses, and provisioning
- Credit deterioration as of December 2020:
  - About one-fifth of credit portfolios were either non-performing or restructured, representing about 10 percent of GDP.
  - Share of restructured loans: 19 percent of total loans in December 2020; 20½ percent of total loans in May 2021.
  - Magnitude of restructured loans equals about 11 percent of GDP (systemic perspective).
  - Around 2 percent of loans were under moratoria with overlapping expirations until July 2021.
  - NPLs stand at around 8 percent of gross loans and are higher for FX-denominated loans.
- Distribution of losses by portfolio and sector:
  - Largest losses originate from retail portfolios, and from FX loans to SMEs and corporates.
  - Within retail: FX-denominated mortgages and lari consumer loans likely to pose substantial losses.
  - Within corporate and SME books: hotels, restaurants, consumer durables, construction, and real estate services expected to be heavily impacted.
- C-VaR 1-year 99 percent results (as percent of gross exposures / performing loans):
  - Expected Loss to Gross Exposure: 2.3 (2020), 2.7 (Baseline), 4.1 (Extended Pandemic), 6.7 (Capital Outflows).
  - VaR to Gross Exposure: 6.8 (2020), 7.0 (Baseline), 10.4 (Extended Pandemic), 17.0 (Capital Outflows).
  - Expected credit losses in 2021: Baseline about 2¾ percent of gross exposures; Extended Pandemic about 4 percent; Capital Outflows about 6¾ percent.
  - Tail losses (99th percentile) as percent of performing loans: Baseline about 7 percent; Extended Pandemic up to 17 percent.
- Comparison with historic average losses (2012‒2020 adjusted):
  - Historic yearly losses: about 688 mil. lari.
  - Baseline expected losses: 40 percent larger than usual.
  - Extended Pandemic: two times larger than usual.
  - Capital Outflows: up to 3½ higher than usual.
- Loan loss reserves and provisioning:
  - Loan loss reserves: 2,434.0 (Mil. GEL) (memo item).
  - Current loan loss provisioning on large exposures stands at about three times expected losses according to bank PD and LGD estimates.
  - Ratio of loan loss provisions to expected C-VaR losses remains above 100 percent across three scenarios for the majority of banks, except for 5 banks; provision shortage severe only for two small banks.

### Key results — solvency and capital adequacy
- Pre-pandemic/entry position and pandemic impact:
  - Preemptive loan loss provisioning in 2020 reduced reported capital ratios by almost 2 percentage points.
  - Basel CAR was 17.6 percent and Tier 1 capital was 12.8 percent in December 2020.
  - Interest margins stand above 500 basis points.
  - ROA: dropped from an average of about 2½ percent to ‒7 percent in March 2020; recovered for the year to slightly positive levels driven by the two largest banks.
  - Loan to deposit ratio surpasses 120 percent.
  - Dollarization of credit portfolios is still high at around 50 percent.
- Scenario outcomes (three-year horizon):
  - Baseline:
    - Banking system prepared to absorb pandemic losses under the baseline without threatening financial stability.
    - Aggregate CAR drops by 1 p.p. to around 16½ percent in 2021 and starts to recover from 2022 on.
    - Replenishment of bank capital buffers likely to take some 2-3 years.
  - Extended Pandemic:
    - Deeper deterioration of credit quality; three banks requiring fresh capital (FSAP and NBG analyses indicate similar small-bank needs).
  - Capital Outflows:
    - Four small and medium-sized banks require additional capital to meet minimum capital requirement.
    - Aggregate capital shortfall in most severe scenario is about 1½ percent of GDP.
    - CAR falls to 10.8 and Tier 1 to 5.3 in the most extreme scenario (aggregate cursory comparison).
- C-VaR implied system-level metrics (selected):
  - Tier 1 Ratio: 12.7 (2020), 11.5 (Baseline), 9.1 (Extended Pandemic), 5.3 (Capital Outflows).
  - CAR: 17.5 (2020), 16.6 (Baseline), 14.4 (Extended Pandemic), 10.8 (Capital Outflows).
  - VaR (Mil. GEL): 2,464.3 (2020), 3,626.6 (Baseline), 5,935.5 (Extended Pandemic).
  - Expected Loss (Mil. GEL): 959.0 (2020), 1,424.5 (Baseline), 2,346.8 (Extended Pandemic).
  - Total Gross Loans: 38,234.0 (Mil. GEL) (all scenarios).
  - Total Assets: 57,051.0 (Mil. GEL) (all scenarios).
  - Tier I: 6,059.0 (Mil. GEL) (all scenarios).
  - Eligible Total Capital: 8,320.0 (Mil. GEL) (all scenarios).
  - Risk-Weighted Assets: 47,539.7 (Mil. GEL) (all scenarios).

### Liquidity assessment and interbank contagion
- System liquidity:
  - The system has enough high-quality liquid assets to endure severe deposit withdrawals (both in lari and FX) during a month while maintaining regular operations.
  - All banks were able to endure extreme shocks to their liabilities during a month with overall LCR ratios above 100 percent even after restricting gross inflows from loans and securities to zero.
  - Lari liquidity tighter: the lari LCR of one small bank dropped below 75 percent.
  - Complementary exercise excluding SME collateral pledged for central bank liquidity resulted in one small bank slightly failing the LCR test in lari; overall LCR remained comfortable at 120 percent.
- Funding and rollover risks:
  - Wholesale funding is mainly long-term and includes financing from parent companies and IFIs.
  - Structural lari liquidity shortage and reliance on central bank facilities remain areas to monitor.
  - In May 2021 the LCR requirement in lari was restored.
  - Authorities plan to gradually unwind the eligibility of SME loans as collateral for central bank liquidity over two years starting at the end of 2021.
- Interbank contagion:
  - Gross interbank loans represent only 0.5 percent of total bank assets.
  - Matrix of interbank exposures is sparse; contagion immaterial from systemic perspective in the model.
  - Cursory failure of an individual bank produces a 2.2 p.p. average drop in CAR ratios among lenders and a post-shock minimum CAR of 12.6 percent.
  - Contagion caveat: confidence effects from the failure of a D-SIB could still transmit systemic stress; two largest banks dominate the Contagion Index.

### Corporate stress tests and large exposures
- Sample and coverage:
  - Corporate sample comprised financial information for 114 firms on a solo basis and 49 consolidated financial statements.
  - Assets of sampled firms added up to about 28 percent of GDP on a consolidated basis.
  - Total loans to the sampled borrowers represented about 16 percent of total bank loans and 72 percent of banks’ capital on average.
  - Nominal GDP in 2020: 49,888 Mil. GEL.
- Corporate financial health findings:
  - About 45 percent of firms analyzed (representing 44 percent of sampled assets) displayed at least one weak financial ratio at the consolidated level.
  - Companies with ICR below 100 percent accounted for about 12 percent of sampled assets.
  - Companies with DEBITDA above 6 represented 24 percent of sampled assets.
- Exposure concentration and collateral:
  - Banks’ ten largest exposures add up to 11 percent of GDP and account for about two-thirds of regulatory capital (with dispersion across banks).
  - Four small institutions have large exposures sometimes surpassing 150 percent of Tier 1 capital.
  - Large exposures vis-à-vis natural persons represent about ½ of a percent of GDP and 3½ percent of banks’ Tier 1 capital.
  - Collateral composition for large exposures: real estate collateral ~56 percent of total collateral value; third-party guarantees add ~40 percent.
  - Median ratio of real estate guarantees to loans: 119 percent for FX-denominated loans and 100 percent for lari exposures.
  - Haircuts applied in provisioning: 30 percent to real estate collateral; 10 percent to third-party guarantees.
- Corporate stress test outcomes:
  - Additional provisioning stands at about 2 percent of Tier 1 under the Baseline stress scenario.
  - Stressed scenarios produced benchmark loan loss provisions of about 6 percent of gross large exposures in the most severe Capital Outflows scenario.
  - Additional provisions relative to gross large exposures amounted to around 2 percentage points above current level in the two distressed scenarios.
  - Most additional provisioning appears manageable except for 2 small banks.

### Market, interest rate, sovereign and other sensitivities
- Market risk:
  - Equity price risk immaterial.
  - Interest rate risk: parallel shift in yield curve by 300 bps would lead to an average increase in ROA by 0.05 percentage points; change in ROA of individual banks ranges from ‒0.3 to 0.4 percentage points.
  - A 300 basis point parallel increase in the yield curve would lead to a 3.9 percentage point drop in average CAR.
- FX sensitivity:
  - Net open positions in FX stand at 0.2 percent of regulatory capital on average (range ‒3.3 percent to 15.3 percent across banks).
  - A 20 percent lari depreciation would lead to a drop in the CAR ratio of the bank with the largest net short FX position of about 0.6 percentage points.
- Sovereign-bank feedback loop:
  - Banks’ holdings of sovereign bonds are held to maturity and sovereign debt remains sustainable.
  - Even a 600 bps shock to sovereign spreads could be absorbed without material impact on banks’ solvency if bonds were marked-to-market.

### Key policy recommendations (selected)
- Enforce retention of earnings until pandemic-related uncertainties dissipate, credit losses are absorbed and capital buffers are fully restored — ST.
- Implement concentration limits on large exposures to Tier 1 capital as planned — ST.
- Continue work to improve the availability and quality of data on large borrowers — ST.
- Strengthen data collection on FX hedging of corporate borrowers — I.
- Continue to ensure accurate and transparent classification of restructured loan portfolios throughout the pandemic.
- Continue to improve the credit information environment, including use of the forthcoming Credit Registry planned to be launched by end-2021, and enhance reporting of corporate FX revenues/expenses and non-bank foreign currency debt.

_Source: EXECUTIVE SUMMARY and selected chapters, 1geoea2021005 (NBG and IMF staff calculations; IMF FSAP team)._

### EXECUTIVE SUMMARY __________________________________________________________________________________ 5

### EXECUTIVE SUMMARY

### Background and system structure
- Banking system size and concentration
  - Total bank assets are about 115 percent of GDP.
  - The three largest banks account for about 85 percent of total assets.
  - 14 out of the 15 banks have some foreign ownership; there are no public banks.
- Financial dollarization and external exposure
  - About half of bank loans and two-thirds of bank liabilities are denominated in FX.
  - Non-resident deposits account for about one-fourth of retail deposits and 7½ percent of banks’ balance sheets.
  - International reserves stand at adequate levels according to standard metrics, but persistent current account deficits have led to the accumulation of substantial foreign liabilities.
- Funding structure and liquidity
  - As of end-December 2020, retail deposits represent about 36 percent of total liabilities; deposits from non-financial companies add an additional 32 percent.
  - The share of FX deposits in total deposits has remained slightly above 60 percent.
  - The system displays a structural shortage of lari liquidity; banks partially roll over short-term central bank funding.
- Macroeconomic shock from COVID-19
  - The economy contracted by 6.1 percent in 2020.
  - Nominal FX depreciated by 17.7 percent vis-à-vis the USD in March 2020; the year closed with a more moderate average nominal depreciation of about 10 percent.
  - The unemployment rate increased to about 18½ percent.
  - Fiscal deficit ballooned to 8.9 percent of GDP and public debt jumped from about 40 percent of GDP in 2019 to 60 percent of GDP in 2020.

### Stress test scope, methodology, and focus
- Exercises and implementation
  - Top-down stress tests carried out jointly by the NBG and the FSAP teams during January‒April 2021.
  - Complementary bottom-up exercises were not implemented at the authorities’ request due to COVID-19 operational challenges.
- Primary focus areas
  - Credit and liquidity risks, with particular emphasis on FX-induced credit risk and concentration risk from lending to large corporates or sectoral clusters.
  - Liquidity assessment included currency-specific LCR analysis and potential sudden calls on liabilities, including by non-residents.
- Scenarios and diagnostics
  - Baseline, two low-probability but plausible distressed scenarios: an Extended Pandemic with sluggish recovery, and a Capital Outflows scenario combining the extended pandemic with an adverse external financial environment.
  - Sensitivity analyses on market, interest rate, and FX shocks; liquidity shocks calibrated using dynamics at the peak of the 2008‒09 global financial crisis.

### Key results — credit quality and losses
- Current credit deterioration
  - As of December 2020, about one-fifth of credit portfolios were either non-performing or restructured, representing about 10 percent of GDP.
  - Loan restructuring widespread; deeper systemic materiality in tourism-related segments, mortgages, and real estate activities.
  - Impact more severe for FX-denominated loans, highlighting FX-induced credit risk from high financial dollarization.
- Distribution of losses by portfolio and sector
  - Largest losses originate from retail portfolios, and from FX loans to SMEs and corporates.
  - Within retail: FX-denominated mortgages and lari consumer loans likely to pose substantial losses.
  - Within corporate and SME books: hotels, restaurants, consumer durables, construction, and real estate services expected to be heavily impacted.
- Data constraints
  - Scope exists to improve data on natural FX hedging of borrowers based on FX income.
  - NBG’s credit registry initiative targeted for completion by end-2021 to help data collection.

### Key results — solvency and capital adequacy
- Baseline scenario
  - Banking system prepared to absorb pandemic losses under the baseline without threatening financial stability.
  - Credit losses can be absorbed with current capital cushions and internal capital generation, though it will take a few years for banks to restore pre-crisis capital buffers.
  - Authorities encouraged to enforce retention of banks’ earnings until pandemic-related uncertainties dissipate and capital cushions are fully restored.
- Distressed scenarios (three-year horizon)
  - System appears resilient but capital injections would be needed in some banks under stressed scenarios.
  - In the most severe scenario, four small and medium-sized banks require additional capital to meet the minimum capital requirement.
  - Aggregate capital shortfall in this most severe scenario is about 1½ percent of GDP.
  - Stress tests implemented by the NBG produced somewhat lower loss estimates; results broadly consistent with FSAP team findings.

### Liquidity assessment
- High-quality liquid assets and LCR
  - The system has enough high-quality liquid assets to endure severe deposit withdrawals (both in lari and FX) during a month while maintaining regular operations.
  - Liquidity stress shocks based on 2008‒09 peak dynamics.
  - A complementary exercise excluding SME collateral pledged for central bank liquidity resulted in one small bank slightly failing the LCR test in lari; overall LCR remained comfortable at 120 percent.
- Funding and rollover risks
  - Wholesale funding is mainly long-term and includes financing from parent companies and IFIs.
  - Structural lari liquidity shortage and reliance on central bank facilities remain areas to monitor.

### Other risks and sensitivities
- Sovereign-bank feedback loop
  - Not a material concern: banks’ holdings of sovereign bonds are held to maturity and sovereign debt remains sustainable.
  - Even a 600 bps shock to sovereign spreads could be absorbed without material impact on banks’ solvency if bonds were marked-to-market.
- Market and interest rate risks
  - Equity price risk is immaterial.
  - Interest rate risk from a parallel shift in the yield curve by 300 bps and direct FX risk from an additional 20 percent depreciation in the lari can be absorbed without threatening solvency of individual banks.

### Corporate stress tests and large exposures
- Corporate borrower financial health
  - Financial ratios of larger corporate borrowers broadly sound on aggregate, with dispersion across firms.
  - About 8 percent of companies analyzed (representing 12 percent of assets) appeared unable to cover their interest expenses.
- Concentration risk and large exposures
  - On average, the ten largest exposures of each bank represent about two-thirds of Tier 1 capital.
  - Corporate stress tests integrated into bank solvency analysis indicate concentration risk does not threaten financial stability.
  - Moderate increase in provisions on large exposures under adverse scenarios suggested; most banks, except two small ones, should absorb these without threatening solvency.
- Basel large exposure limit
  - Implementation of the Basel prudential limit on large exposures would be binding for only a handful of banks and firms.
  - Estimated impact on the supply of bank credit to large corporates would be small.
  - Authorities are encouraged to implement the new regulation on large exposure limits following Basel standards as planned.

### Key recommendations (Table 1: Georgia: Summary of Key Recommendations)
- 1 Enforce retention of earnings until pandemic-related uncertainties dissipate, credit losses are absorbed and capital buffers are fully restored — ST
- 2 Implement concentration limits on large exposures to Tier 1 capital as planned — ST
- 3 Continue work to improve the availability and quality of data on large borrowers — ST
- 4 Strengthen data collection on FX hedging of corporate borrowers — I

Notes:
- Authority responsible for implementation is assumed to be the NBG.
- Timing: I; immediate (<1 year); ST; short term (1–3 years); MT: medium term (3–5 years).

*Source: EXECUTIVE SUMMARY, 1geoea2021005*

### 6.      The banking system started the pandemic from a strong position. Banks entered the

### 6. The banking system started the pandemic from a strong position. Banks entered the

### Banking system initial strengths and pandemic response
- Banks entered the crisis with strong profitability, liquidity, and solvency indicators driven by strict prudential requirements and buildup of capital and liquidity buffers.
- Banks adapted quickly to operate in a virtual environment and voluntarily accommodated massive moratoria requests from borrowers; successive waves of moratoria involved smaller loan volumes as some borrowers resumed repayments and others restructured loans.
- Preemptive loan loss provisioning in 2020 reduced reported capital ratios by almost 2 percentage points.
- Basel CAR was 17.6 percent and Tier 1 capital was 12.8 percent in December 2020.
- Interest margins stand above 500 basis points, supporting loss-absorption capacity.
- Trading income accounted for only 11 percent of total income; banks tend to hold bond holdings to maturity and bank positions on derivatives are negligible.
- Short-term liquidity indicators dropped in March 2020 but recovered afterwards with policy support; structural liquidity is more challenging with loan-to-deposit ratios surpassing 120 percent.
- Dollarization of bank balance sheets is substantial, generating FX-induced credit risk and liquidity risk; banks maintain closed FX positions within prudential limits so direct FX risk is not material.

### Credit quality and non-performing / restructured loans
- Non-performing loans according to the customary 90-day past due definition remained flat during 2020, but the NBG’s more conservative/forward-looking definition shows substantial deterioration in credit quality driven by downgrades of restructured loans.
- Share of restructured loans:
  - 19 percent of total loans in December 2020.
  - 20½ percent of total loans in May 2021.
- Magnitude of restructured loans equals about 11 percent of GDP (systemic perspective).
- Around 2 percent of loans were under moratoria with overlapping expirations until July 2021.
- NPLs stand at around 8 percent of gross loans and are higher for FX-denominated loans.
- NPL averages mask wide dispersion across banks; a couple of small banks have sizable NPL shares; large share of NPLs are restructured loans reclassified to higher risk categories.

### Key financial soundness indicators (selected, as reported)
- Bank capital ratios dropped by almost 2 percentage points in 2020 due to preemptive provisioning.
- ROA: dropped from an average of about 2½ percent to ‒7 percent in March 2020; recovered for the year to slightly positive levels driven by the two largest banks.
- Basel CAR (Dec 2020): 17.6 percent.
- Tier 1 capital (Dec 2020): 12.8 percent.
- Trading income share of total income: 11 percent.
- Loan to deposit ratio: surpasses 120 percent (structural liquidity challenge).
- Dollarization: text notes dollarization of credit portfolios is still high at around 50 percent.

### Stress test scope and implementation
- Coverage: bank-level assessment covering all fifteen banks; three largest banks jointly account for about 85 percent of system assets and are systemically important.
- Approach: coordinated top-down stress tests by the NBG and FSAP teams using supervisory data, an in-depth 2020 loan quality assessment, and dedicated data requests to banks. Bottom-up stress tests were not carried out at the NBG’s request due to operational difficulties facing banks.
- Components assessed:
  - Corporate stress tests focused on the ten largest exposures of individual banks (exposures included loans to related firms within conglomerate groups); results matched with lender banks to assess adequacy of loan loss provisions and fed into credit risk models.
  - Credit risk: assessed FX-induced credit risk and concentration risks; loan book split into corporate, SME, retail, and large exposures; corporate/SME split into 32 categories; retail split into mortgages, consumer loans, and other retail loans.
  - Market risks: focused on interest rate risk (including banking book) and FX risk; equity price risk considered immaterial.
  - Sovereign risk: bank holdings of sovereign bonds about 10 percent of total bank assets and held to maturity; sovereign risk assessed via impact of widening sovereign spreads.
  - Liquidity risk: focused on liabilities side given small trading portfolios and substantial dollarization of liabilities and importance of non-resident deposits; data partitioned by currency and residency.
  - Contagion: contagion analysis used interbank exposures.
- Cut-off date: December 2020.

### Stress test methodologies and parameters
- Stress tests leveraged NBG monitoring tools, a recent Technical Assistance on FX-induced credit risk, and in-house IMF models on credit and liquidity risk.
- Hurdle rates for solvency analysis:
  - Basel CAR: 8 percent.
  - T1: 6 percent.
  - CT1: 4.5 percent.
- Liquidity hurdle rates (LCR-based):
  - Overall LCR requirement: 100 percent.
  - Currency-specific LCR: 100 percent for FX, 75 percent for lari.
- Methodological assumptions:
  - Semi-static bank balance sheets: bank loans restricted to grow at the pace of economic activity in each stress scenario.
  - Interest rate risk assessed on entire balance sheet and relevant off-balance positions, accounting for marked-to-market valuation and repricing gap effects.
  - FX risk assessed on entire balance sheet without accounting for cross-currency changes.
  - Credit risk assessment used scenario analysis, credit VaR, and sensitivity analysis; leveraged detailed partition of loan portfolios (32 credit types and ten largest exposures, and two currencies) to capture FX-induced credit risk and concentration.
  - Liquidity stress tests based on cash flow analysis building on LCR methodology.

### Stress test scenarios (three-year horizon 2020–22)
- Scenarios span three years 2020‒22; severity measured by GDP deviation relative to baseline.
- Baseline:
  - Based on Spring 2021 WEO projections.
  - Assumes pandemic contained with vaccine distribution and economy starts to recover in second half of 2021.
  - Real GDP growth reached 3½ percent in 2021 and reached 2019 levels by mid-2022; in outer years GDP growth quickly returns to pre-pandemic dynamics.
  - Baseline may prove conservative given faster-than-envisaged rebound in early 2021.
- Scenario 1: Extended Pandemic
  - Pandemic extends into late 2021 due to virus mutations, vaccine distribution lags, and population resistance.
  - Afterwards, sluggish growth from phased withdrawal of support measures and lower potential due to scarring and long-lasting effects on production processes.
  - Assumes continuation of a benign external financial environment with reserve-currency countries maintaining expansionary fiscal and unconventional monetary policies.
- Scenario 2: Capital Outflows
  - More severe and less plausible; similar pandemic progression plus adverse external financial environment.
  - Captures risks from large stock of external liabilities, high dollarization, and sizable gross external financing needs.
  - Triggers could include political turmoil, concerns on solvency of a systemic institution, or increase in global risk aversion.
  - Outcomes: sudden stop or capital outflows, currency depreciation, spike in interest spreads, and a defensive hike in the policy rate with adverse feedback effects on activity.

### Scenario calibration and mapping to credit sectors
- Calibration combined statistical analysis and qualitative judgment due to structural nature of the COVID shock and limited historical relevance.
- Applied a (smaller-than-usual) 1–1.5 standard deviations below the baseline over the first two years.
- Sectoral projections were produced using time series models and mapped into 25 commercial credit types used by the NBG to simulate pandemic impacts on borrower activity.
- Corporate stress test sample targeted ten largest exposures per bank, using latest available financial statements provided by lender banks.

_Source: NBG and IMF staff calculations; National Bank of Georgia; IMF FSAP team._

### 16.      Credit risk stemming from large corporate borrowers was assessed using company

### 16. Credit risk stemming from large corporate borrowers was assessed using company financial data provided by the lender banks

### Sample and data coverage
- Sample comprised financial information for 114 firms on a solo basis and 49 consolidated financial statements.
- Assets of sampled firms added up to about 28 percent of GDP on a consolidated basis.
- Total loans to the sampled borrowers represented about 16 percent of total bank loans and 72 percent of banks’ capital on average.
- Table 5 (summary by lender bank): Total consolidated assets 19,188 Mil. GEL (38.5 Pct. GDP) for 60 consolidated statements; Total unconsolidated assets 20,561 Mil. GEL (41.2 Pct. GDP) for 132 unconsolidated statements. Total Adjusted (excluding common borrowers across banks): 13,886 Mil. GEL (27.8 Pct. GDP) for 49 consolidated; 12,476 Mil. GEL (25.0 Pct. GDP) for 114 unconsolidated.
- Nominal GDP in 2020: 49,888 Mil. GEL.
- Note: latest available unconsolidated financial statements (2018-19).

### Methodology for corporate financial assessment
- Used lender banks’ recent financial statements for each borrower firm; firms with insignificant amounts of bank loans were assessed by proxying with their conglomerate group financial situation.
- Analysis exploited consolidated data where available; many assessments relied on unconsolidated data when consolidated statements were not available.
- Financial ratios analyzed (customary set measuring liquidity, profitability, leverage, solvency): short-term debt to total debt; debt to EBITDA; return on assets (ROA); return on equity (ROE); quick ratio; current ratio; interest coverage ratio (ICR); debt to earnings before interest and taxes adjusted by depreciation and amortization (DEBITDA); debt to equity (DE).
- Binary vulnerability thresholds: negative equity to assets; negative ROA; short-term debt above 50 percent of total debt; ICR below 10 percent. Composite indicator ranges in [0‒4] (larger = weaker).

### Data limitations and interpretation caveats
- Analysis based on data provided under a tight FSAP schedule; consolidated statements at the conglomerate group level not available in all cases.
- Each borrower firm had either consolidated or unconsolidated statements (not both); identification of companies included in parents’ consolidated statements at sub-group levels was not always possible.
- Corporate solvency analysis is static and backward-looking and cannot account for company prospects, shareholder support, or qualitative information used in supervisory assessments.

### Exposure concentration and related-party lending
- Banks’ ten largest exposures add up to 11 percent of GDP and account for about two-thirds of regulatory capital (with dispersion across banks).
- Four small institutions have large exposures sometimes surpassing 150 percent of Tier 1 capital.
- Large exposures vis-à-vis natural persons represent about ½ of a percent of GDP and 3½ percent of banks’ Tier 1 capital.
- Lending to related parties is not prevalent and mainly comes from five banks; only one small bank appears relatively more exposed to concentration risk and related-party lending, but amounts not of systemic concern.

### Corporate financial health findings
- About 45 percent of the number of firms under analysis (representing 44 percent of the sampled assets) displayed at least one weak financial ratio at the consolidated level (negative equity, negative/weak profits, high leverage, reliance on short-term debt).
- Companies with ICR below 100 percent accounted for about 12 percent of the sampled assets.
- Companies with DEBITDA above 6 represented 24 percent of the sampled assets.
- Distributions of ratios have long tails; some firms display multiple weaknesses.
- Binary-indicator analysis suggests only a few small banks have more than 75 percent of their capital compromised by large exposures with weak financials (see Table 7).

### Collateral coverage
- Large exposures generally well collateralized: real estate collateral ~56 percent of total collateral value assessed; third-party guarantees add ~40 percent.
- Other collateral types include bank deposits, gold, and state guarantees.
- Median ratio of real estate guarantees to loans: 119 percent for FX-denominated loans and 100 percent for lari exposures.
- Median coverage of third-party guarantees: 14 percent for FX loans and negligible for lari loans.
- Cash and gold taken at full value; government guarantees taken at full value.
- Haircuts applied in provisioning analysis: 30 percent haircut to real estate collateral; 10 percent haircut to third-party guarantees.

### Loan loss provisioning and PD/LGD observations
- Current loan loss provisioning on large exposures stands at about three times expected losses according to bank PD and LGD estimates.
- Banks provided very dissimilar PD estimations for common borrowers; distribution of banks’ PD estimates appeared unrelated to firms’ exposure to COVID-sensitive activities.
- Bank provisioning methodology follows NBG’s methodology based on borrowers’ financial conditions.

### Benchmark provisioning methodology (FSAP team)
- Net loan exposures = gross exposures (loan amounts + off-balance items) minus collateral (after haircuts).
- Haircuts: real estate 30 percent; third-party guarantees 10 percent (except government guarantees at full value); cash and gold full value.
- For non-financial companies, benchmarks estimated by comparing ICR and DEBITDA with values in Table 8.
- Benchmark provisions for companies with negative equity to assets set at 100 percent.
- Financial companies assessed by capital to asset ratios rather than ICR/DEBITDA.
- Table 8 (Loan Loss Provisions by intervals of ICR and DEBITDA): provisions vary by DEBITDA and implied DSTI/ICR/equity to assets buckets (examples from table):
  - DEBITDA < 4.5 / DSTI < 52 / ICR < 120 / EA > 14 => Loan Loss Provisions 2
  - 4.5 < DEBITDA < 6.0 / 52 < DSTI < 69 / 100 < ICR < 120 / 8 < EA < 14 => Loan Loss Provisions 10
  - 6.0 < DEBITDA < 8.5 / 69 < DSTI < 97 / 70 < ICR < 100 => Loan Loss Provisions 30
  - 8.5 < DEBITDA < 11.5 / 97 < DSTI < 132 / 50 < ICR < 70 => Loan Loss Provisions 50
  - 11.5 < DEBITDA / 132 < DSTI / ICR < 50 / EA < 8 => Loan Loss Provisions 100
- Note: Implied DSTI based on median effective interest rate paid by large corporations (7.7 percent) and assuming a fifteen year maturity.

### Stress test results and implications
- Corporate stress tests used the same bank stress test scenarios, augmented with the actual COVID shock in 2020 (corporate data were pre-pandemic).
- Classification by economic activity used projected sector evolution during 2020–23 to adjust ICR and EBITDA; shocks to interest rates and lari/USD exchange rate were applied per stress scenarios.
- Additional provisioning stands at about 2 percent of Tier 1 under the Baseline stress scenario (Table 9).
- Stressed scenarios produced benchmark loan loss provisions of about 6 percent of gross large exposures in the most severe Capital Outflows scenario.
- Additional provisions relative to gross large exposures amounted to around 2 percentage points above current level in the two distressed scenarios.
- Most additional provisioning appears manageable except for 2 small banks.
- Robustness check shortening stress horizon to 2020–21 (censoring rebound) yielded similar qualitative results.
- Lack of sensitivity to stress scenarios suggests many weaker companies’ reference financial ratios do not migrate to lower bucket intervals under these scenarios.

*Source: NBG and Fund staff calculations.*

### 26.      The NBG  has a comprehensive monitoring of the large corporate borrowers, but the

### 1geoea2021005 - 26.      The NBG  has a comprehensive monitoring of the large corporate borrowers, but the 

### Monitoring of large corporate borrowers and data gaps
- The NBG tracks large borrowers closely and uses both financial data and soft information to assess adequacy of loan loss provisions on large exposures.
- Banks are required by regulation to maintain up-to-date financial information on their 100 largest debtors; this is challenging because most firms are not listed.
- Information reported by banks during the FSAP had generally good coverage, but some data inconsistencies indicate processes to collect, clean, and systematize data could be improved.
- Data gaps identified:
  - Limited information to determine exposure to FX-risk of large corporate borrowers, specifically on FX revenues and expenses, and FX-denominated debt issued outside the domestic banking system.
  - Financial ratios for companies without financial statements were estimated by taking weighted averages of all companies within the same conglomerate group.
- Recommended improvements:
  - Enhance data collection through lender banks.
  - Share non-sensitive firm data on exports and imports between government agencies.
  - Use the forthcoming credit registry and related reporting forms to strengthen data availability.

### Single-party concentration limits (large exposure regulation)
- The planned regulation sets single-party concentration limits of 25 percent of Tier 1 capital, in line with Basel standards.
- Authorities are concerned about potential impacts on banks’ ability to finance large projects given high dependence of Georgian firms on bank lending and shallow capital and money markets.
- Loan syndication offers limited alternative due to high concentration in the Georgian banking system.
- Forward-looking hypothetical simulation: increasing the size of large exposures from 20 to 60 percent from current levels.
  - Results: the limit would bind for about 30 percent of the firms under analysis.
  - Estimated macro impact: maximum loan shortage around 2 percent of GDP.
- Table 10 (summary results):
  - Total number of firms that would hit limits if loan demand increases by:
    - Actual: 8
    - 20: 13
    - 40: 21
    - 60: 33
  - Estimated loan shortage, in percent of GDP (Total):
    - Actual: 0.2
    - 20: 0.6
    - 40: 1.2
    - 60: 1.9
- Conclusion: enforcing the single-party limit should not induce a sizable constraint on banks’ capacity to provide credit to the economy.

### Bank solvency assessment: approach and challenges
- The assessment used multiple approaches due to uncertain pandemic effects on credit losses and policy responses.
- Challenges:
  - Atypical nature of the COVID shock weakened usefulness of historic data to assess credit losses.
  - Government support measures reduced vulnerabilities, concealing underlying financial conditions.
  - Loan moratoria and restructurings masked underlying effects on credit portfolio performance.
- Approach taken:
  - FSAP leaned towards conservative estimations of bank losses and profits in projected scenarios.
  - Risk parameters estimated using pre-COVID data with uncertainty around point estimates included where possible.
  - Sensitivity analyses were used to inform the solvency assessment.

### Credit Value at Risk (C-VaR) model and inputs
- C-VaR provided initial evaluation of credit risk and adequacy of existing loan loss provisions and capital.
- Granular partitioning of credit portfolios:
  - Four segments: corporate, SME, Retail, and Large Exposures.
  - Corporate and SME split into 25 categories following NBG practice.
  - Retail split into 7 categories (including housing mortgages and consumer loans).
  - Large exposures used inputs from the corporate stress test: posted collateral, main economic activity, bank estimates of PDs and LGDs.
- Model encompassed 39 credit types in two currencies → resulting in 78 categories.
- Estimation of credit risk parameters:
  - PDs estimated from historic data on nonperforming loan flows and stocks at bank level during 2012‒20.
  - Estimated PDs ranged between 3‒7½ percent.
  - Lower PDs for mortgages (no significant differences between FX and lari mortgages).
  - Higher PDs for FX-denominated corporate loans and lari loans to SMEs, with relatively large standard deviations.
  - LGD calibration:
    - Based on market value of collateral with a 30 percent haircut.
    - Overcollateralized exposures subject to an LGD floor of 10 percent.
    - Uncollateralized transactions assigned a flat 45 percent LGD.
    - Rationale: conservative given Georgian legal framework allows full recourse against defaulting borrowers.

### PD estimation details (Box 1) — numeric PDs by loan type and currency (2012-20)
- Mean and St. Dv. of PDs by Loan Type:
  - GEL Loans
    - Consumer: Mean 0.046, St. Dv. 0.049
    - Corporate: Mean 0.050, St. Dv. 0.085
    - Mortgages: Mean 0.029, St. Dv. 0.071
    - Retail: Mean 0.043, St. Dv. 0.047
    - SME: Mean 0.069, St. Dv. 0.120
  - FX Loans
    - Consumer: Mean 0.041, St. Dv. 0.058
    - Corporate: Mean 0.074, St. Dv. 0.164
    - Mortgages: Mean 0.030, St. Dv. 0.040
    - Retail: Mean 0.044, St. Dv. 0.116
    - SME: Mean 0.046, St. Dv. 0.067

### Mapping macroeconomic conditions to PDs (Box 2)
- Satellite panel econometric model maps macro variables to logit(PDs) using panel data of credit types × banks for 2012‒20.
- Specification included lagged PDs, unemployment rate, and lagged GDP growth; FX regressions also included lagged average FX depreciation.
- Key estimated quantitative sensitivities:
  - A 1 percentage point increase in the unemployment rate → about 1 percentage point increase in average PDs over one year.
  - A 10 percentage point increase in FX depreciation → about 0.8 percentage point increase in average PDs over one year.
- Regression table excerpt (panel regressions of logit of PDs):
  - Lagged Logit of PD: 0.1218 (FX), 0.1251 (GEL)
  - Unemployment Rate: 0.3524 (FX), 0.2945 (GEL)
  - Lagged GDP Growth: -0.1654 (FX), -0.1042 (GEL)
  - Lagged FX Rate Depreciation: 0.0266
  - Dummy Mortgage: -0.2754 (FX), -0.6379 (GEL)
  - Constant: -7.6648 (FX), -6.4784 (GEL)
  - Observations: 167 (GEL regression) and 235 (FX regression)
  - Number of Panels: 49 (GEL), 59 (FX)
  - R-Squared: 0.17 (FX), 0.19 (GEL)

### C-VaR results and implications
- Expected credit losses in 2021:
  - Baseline: about 2¾ percent of the gross exposures.
  - Extended Pandemic: about 4 percent of the gross exposures.
  - Capital Outflows scenario: about 6¾ percent of the gross exposures.
- Tail losses (99th percentile) as percent of performing loans (gross exposures):
  - Baseline: about 7 percent
  - Extended Pandemic: up to 17 percent
  - Statement: in the most adverse event, credit losses would remain below 17 percent with 99 percent confidence.
- Comparison with historic average:
  - Historic yearly losses (2012‒2020), adjusted to end-2020 credit portfolio size: about 688 mil. lari.
  - Relative to usual:
    - Baseline expected losses: 40 percent larger than usual.
    - Extended Pandemic: two times larger than usual.
    - Capital Outflows: up to 3½ higher than usual.
- Aggregate capital impact (cursory comparison, assuming 30 percent loss on existing NPLs and ignoring bank profits):
  - Aggregate CAR would fall from 17.7 percent to 16.6 percent under the Baseline, and to 10.8 percent in the most extreme scenario.
  - Tier 1 would fall from 12.7 to 5.3 percent.
- Conclusion: the system appears able to endure the stress scenarios even under very conservative assumptions.

### Adequacy of provisions and capital at bank level
- Ratio of loan loss provisions to expected C-VaR losses:
  - Remains above 100 percent across three scenarios for the majority of banks, except for 5 banks.
  - Provision shortage is severe only for two small banks.
- Capital cushions:
  - Broadly in line with estimated losses at the 99 percentile except for the same 2 small banks.
- Overall conclusion: C-VaR results indicate the system is well prepared to absorb credit losses from the stress scenarios without imperiling financial stability.

*Source: NBG and Fund staff calculations, as presented in the FSAP chapter.*

### 34.      Current risk-weights  on loans appear  broadly consistent with the results of the C-

### 1geoea2021005 - 34.      Current risk-weights  on loans appear  broadly consistent with the results of the C-

### Summary findings
- Current risk-weights on loans are broadly consistent with results of the C-VaR under the Baseline and Extended Pandemic scenarios.
- Estimated risk-weights on credit portfolios based on the CVaR under the Baseline scenario averaged 0.9 for the entire system, compared with a current risk-weight density or around 1.0.
- Calculated risk-weights under the Extended Pandemic averaged 1.2, with 8 banks falling somewhat short of the reference.
- A stricter comparison using the credit risk parameters of the Capital Outflows scenario produced average risk weights of 2.0.
- Conclusion: current risk weights are deemed broadly adequate under the most plausible scenarios.

### Credit VaR results (one-year 99 percent C-VaR) — key statistics
- Expected Loss to Gross Exposure: 2.3 (2020), 2.7 (Baseline), 4.1 (Ext. Pandemic), 6.7 (Cap. Outflows)
- VaR to Gross Exposure: 6.8 (2020), 7.0 (Baseline), 10.4 (Ext. Pandemic), 17.0 (Cap. Outflows)
- Tier 1 Ratio: 12.7 (2020), 11.5 (Baseline), 9.1 (Ext. Pandemic), 5.3 (Cap. Outflows)
- CAR: 17.5 (2020), 16.6 (Baseline), 14.4 (Ext. Pandemic), 10.8 (Cap. Outflows)

Selected parameters (as reported)
- Performing Exposures (Mil. GEL): 35,016 (2020), 35,012 (Baseline), 35,012 (Ext. Pandemic), 35,012 (Cap. Outflows)
- Number of Loans (Million): 3.2 (all scenarios)
- Number of Banks: 15 (all scenarios)
- Number of Credit Types: 78 (all scenarios)
- Number of Currencies: 2 (all scenarios)

Probability of Default (percent)
- Median: 4.7 (2020); 11.1 (Baseline); 11.1 (Ext. Pandemic); 11.1 (Cap. Outflows)
- Mean: 5.5 (2020); 17.3 (Baseline); 17.3 (Ext. Pandemic); 17.3 (Cap. Outflows)

Recovery Rates (percent)
- Median: 51.9 (2020); 55.0 (Baseline); 55.0 (Ext. Pandemic); 55.0 (Cap. Outflows)
- Mean: 60.3 (2020); 65.1 (Baseline); 65.1 (Ext. Pandemic); 65.1 (Cap. Outflows)

Memo items (in Mil. GEL)
- VaR (Mil. GEL): 2,464.3 (2020), 3,626.6 (Baseline), 5,935.5 (Ext. Pandemic)
- Expected Loss (Mil. GEL): 959.0 (2020), 1,424.5 (Baseline), 2,346.8 (Ext. Pandemic)
- Total Gross Loans: 38,234.0 (all scenarios)
- Loan loss reserves: 2,434.0 (all scenarios)
- Total Assets: 57,051.0 (all scenarios)
- Tier I: 6,059.0 (all scenarios)
- Eligible Total Capital: 8,320.0 (all scenarios)
- Risk-Weighted Assets: 47,539.7 (all scenarios)
- o/w on Credit Portfolios: 39,745.9 (all scenarios)

### Scenario analysis — model, implementation, and high-level outcomes
- Model characteristics:
  - Integrates credit and market risks; quasi-static balance sheet.
  - Credit growth follows GDP growth, adjusted by flows of loan loss provisions and FX valuation changes.
  - Loan book partitioned into corporations and households; each broken down by domestic and FX.
  - Exchange rate effects incorporated for assets and liabilities, including deposits.
  - Simulates evolution of net interest income by projecting interest-sensitive assets and liabilities and respective interest rates.
  - Sovereign risk measured via sovereign spreads; direct FX risk via revaluation of FX assets and liabilities.
  - Solvency analysis not integrated with liquidity stress tests (caveat).

- Implementation:
  - Model run in parallel by the NBG and FSAP teams with alternative credit loss modelling approaches.
  - NBG modeled expected credit losses using projected evolution of NPLs.
  - FSAP used estimated PDs and LGDs for granular portfolios and corporate stress test results.

- High-level outcomes:
  - System is well positioned to absorb losses under stress scenarios without threatening financial stability at the systemic level.
  - Largest losses stem from credit portfolios, particularly FX-denominated loans for corporations, SMEs, and mortgages, and lari-denominated consumer loans.
  - A few small banks fall under minimum CAR and Tier 1 ratios and require fresh capital injections; estimated capital shortfalls are not large from the macrofinancial perspective.

### Results of scenario analysis by FSAP team — specifics
- Under the Baseline scenario:
  - System can absorb losses with current loan loss provisions, operational profits, and capital cushions without threatening financial stability.
  - Aggregate CAR drops by 1 p.p. to around 16½ percent in 2021 and starts to recover from 2022 on.
  - Replenishment of bank capital buffers likely to take some 2-3 years (in line with NBG plan to replenish pillar 2 and systemic buffers starting by end-2021 until 2023).

- Under the Extended Pandemic scenario:
  - Deeper deterioration of credit quality; three banks requiring fresh capital.

- Under the Capital Outflows scenario:
  - Substantially larger credit losses; four banks failing the test.
  - Aggregated capitalization shortfalls relatively small; Tier 1 recapitalization reaching 1.5 percent of GDP in the most adverse case.

- Composition and contributors to losses:
  - Heavier losses originate from FX loans to SMEs and corporates, especially companies in hotels and restaurants and consumer durables.
  - Construction, commercial real estate, and FX-denominated housing mortgages are also heavily impacted.
  - Interest rate risk is a secondary contributor; other market risks negligible.

### Results of scenario analysis implemented by the NBG — methodology and conclusions
- NBG stress testing specifics:
  - Estimation of credit risk based on projection of NPLs by fifteen economic sectors, in domestic and foreign currency.
  - Loan loss provisions assumed as actual sectoral provision coverage rate plus up to 10 percent, depending on sector and currency denomination.
    - Note on provisioning adjustment: If the current provisioning rate is less than 40 percent, an additional 10 percent of loan provisions is added; if between 40‒50 percent, an additional 5 percent is added.
  - Interest rate risk assessed with maturity-adjusted gap analysis; net interest income affected as positions reach repricing time; margins compressed under stress based on historical distribution.
  - Net fees and commission income and net non-interest income assumed to follow nominal GDP growth path.
  - Framework assumes static balance sheets without active bank or supervisor responses.

- NBG conclusions:
  - Financial system entered pandemic well prepared due to pre-pandemic financial policies, capital surcharges, and retained profits.
  - Commercial banks made precautionary loan loss provisions of around 3 percent of RWAs at the start of the pandemic; operational profitability offset losses and banks ended 2020 with a small positive profit.
  - Under the Baseline scenario, aggregate capital adequacy is projected to recover to the pre-crisis level by the end of 2022; some banks may take until end-2023 to fully restore buffers; several major banks may restore buffers in 2021.
  - In the Extended Pandemic scenario, capital ratio of all banks remains above the minimum regulatory threshold; Tier 1 decreases only marginally.
  - In the Capital Outflows scenario, sizable losses in foreign currency portfolios and tourism-linked sectors; three banks would need additional capital. NBG considers capitalization needs insignificant enough not to pose systemic risk; capital adequacy starts to gradually recover after the second year.

### Sensitivity analysis — credit risk exercises
- Exercise 1: Defaults on restructured loans with default rates of 30, 50, 70 percent and a flat 55 percent LGD; pre-impairment profits reduced by lost interest on defaulting loans; no RWA adjustments.
  - Result: system can withstand conservative repayment performance of restructured loans; only two banks fail to meet an 8 percent minimum CAR requirement.

- Exercise 2: COVID-specific stressed PDs — pre-pandemic PDs for COVID-sensitive activities augmented by two-times their standard deviations; LGD flat at 55 percent; pre-impairment profits reduced by lost interest revenues.
  - Result: two small and medium-sized banks fail to meet CAR requirements; aggregate capital shortfall inconsequential from systemic stability perspective.

- Table 13 sample CAR outcomes (selected entries)
  - B1: CAR Before 16.0; After losses on restructured loans 30 -> 15.2; 50 -> 13.2; 70 -> 11.2; After COVID-specific stressed PDs -> 10.3
  - B8: CAR Before 14.2; After losses on restructured loans 30 -> 12.0; 50 -> 11.0; 70 -> 10.1; After COVID-specific stressed PDs -> -3.5
  - B15: CAR Before 83.5; After losses on restructured loans 30 -> 84.8; 50 -> 84.5; 70 -> 84.2; After COVID-specific stressed PDs -> 81.3

### Sensitivity analysis — market and other risks
- Market risk findings:
  - Equity price risk is absent.
  - Interest rate risk from repricing gaps is small.
  - A parallel increase in the yield curve by 300 bps would lead to an average increase in ROA by 0.05 percentage points; change in ROA of individual banks ranges from ‒0.3 to 0.4 percentage points.
  - A 300 basis point parallel increase in the yield curve would lead to a 3.9 percentage point drop in average CAR (interest rate risk in the banking book deemed moderate).
  - Net open positions in FX stand at 0.2 percent of regulatory capital on average, ranging from ‒3.3 percent to 15.3 percent of capital across individual banks.
  - A 20 percent lari depreciation would lead to a drop in the CAR ratio of the bank with the largest net short FX position of about 0.6 percentage points.
- Sovereign risk and FX sensitivity tests presented in charts indicate limited effects for most banks.

### Liquidity stress testing note
- The solvency analysis is not integrated with liquidity stress tests; this is noted as a caveat of the exercises.

*Source: NBG and IMF Staff calculations, as presented in the provided content.*

### 50.      Banks have sufficient high-quality liquid assets to survive severe shocks to their

### 1geoea2021005 - 50.      Banks have sufficient high-quality liquid assets to survive severe shocks to their deposit base.

### Liquidity stress test findings
- All banks were able to endure extreme shocks to their liabilities during a month with overall LCR ratios above 100 percent while sustaining their regular operations, even after restricting the gross inflows from loans and securities to zero.
- The shocks were based on the dynamics at the peak of the 2008‒09 global financial crisis.
- Across currencies, liquidity in lari proved tighter:
  - The lari LCR of one small bank dropped below 75 percent (Figure 10).
  - A complementary exercise, taking out the SME collateral pledged for central bank liquidity, resulted in one small bank slightly failing the LCR test in lari, although the overall LCR remained comfortable at 120 percent.
- Policy actions and timing reported in the source:
  - In May 2021 the LCR requirement in lari was restored.
  - Authorities are planning to gradually unwind the eligibility of SME loans as collateral for central bank liquidity over two years starting at the end of 2021.
- Conclusion: The banking system appears ready for the gradual removal of liquidity support measures implemented early in the pandemic.

### Interbank contagion and systemic interconnectedness
- The interbank market is shallow and interbank contagion immaterial from the systemic perspective.
- Key metrics and observations:
  - The matrix of interbank exposures is sparse.
  - Most banks split their interbank deposits between the two largest D-SIBs, but the amounts are relatively small.
  - Gross interbank loans represent only 0.5 percent of total bank assets.
  - A cursory assessment of the effects of individual bank failures on the CAR of lender banks produces a 2.2 p.p. average drop in CAR ratios, and a post-shock minimum CAR of 12.6 percent.
  - There are no bank failures induced directly (or indirectly) by the failure of individual banks.
- Contagion caveat:
  - Contagion could still occur due to confidence effects stemming from the failure of a D-SIB.
  - The two largest banks pose the largest contagion risk in the system:
    - The Contagion Index shows a clear dominance of the two largest D-SIBs.
    - The Vulnerability Index indicates the most vulnerable institutions tend to be the smallest in terms of assets.

### Climate stress test assessment
- Data availability:
  - There was limited data available for an assessment of climate risk for the financial sector.
  - Discussions with the authorities and World Bank suggested limited availability of relevant data to carry out an assessment of banking sector impact from physical climate risks.
- Potential future scope:
  - Possible future climate transition risk analysis using emissions data for different corporate industries along with estimates of possible carbon taxes on greenhouse gases.
  - Rationale: Georgia is a signatory to the Paris Climate Accord of 2016; such analysis could estimate compliance cost for corporates and related effects on their ability to service bank debt.

### Key policy recommendations (from RECOMMENDATIONS)
- Enforce retention of earnings until the pandemic dissipates, credit losses are absorbed, and capital cushions fully restored.
  - Rationale: Resilience rests on solid pre-pandemic capital buffers; banks should fully deploy profit margins to absorb credit losses and rebuild capital.
  - The proposed timetable to replenish capital appears reasonable, but the uncertain speed of recovery calls for prudence and flexibility.
- Implement large exposure limits as planned, following Basel standards.
  - Considerations:
    - Concerns about impact on loan supply are legitimate given corporate dependence on bank financing and large investment needs.
    - FSAP analysis suggests enforcing single exposure limits as per Basel standards would be binding for very few firms and banks.
    - A hypothetical increase of individual large exposures by up to 60 percent does not appear to imply severe restrictions on banks' capacity to provide financing when measured by number of companies affected and against GDP.
    - There may be room to resort to loan syndication, though large concentration in the banking system poses limits to practical use of syndication.
- Continue to ensure accurate and transparent classification of restructured loan portfolios throughout the pandemic.
  - Maintain high supervisory standards on loan classification and disclosure to support market discipline and incentives for banks’ management.
- Continue to improve the credit information environment.
  - The ongoing Credit Registry project planned to be launched by end-2021 should prove extremely useful.
  - Specific areas for improvement:
    - Information to determine the extent of FX-hedging of large corporate borrowers:
      - Corporate financial data compiled by banks and transmitted to the NBG should aim at including revenues and expenses in FX and a partition of non-bank debt in domestic and foreign currency.
      - Establish protocols for data sharing on firm exports and imports between relevant government agencies and the NBG.
    - Organization of financial data for analytical purposes:
      - Classification of financial statements within corporate groups should seek to replicate conglomerate ownership structures, allowing tracing individual companies to consolidated statements of their parents at the sub-group or group levels.
    - Data quality:
      - Improve the accuracy and completeness of data compiled and processed by banks due to limited public information on company financials.

### Appendix I — Stress Testing Matrix (selected assumptions and parameters)
- Banking sector solvency stress test institutional perimeter:
  - All banks (fifteen institutions with total assets equivalent to about 100 percent of GDP), including the three D-SIBs.
  - Market share: About 95 percent of total financial system assets.
  - Baseline date: Balance sheets as of October 2020, with a later update to target December 2020 as the final cutoff date.
- Methodology and horizon:
  - Two credit risk methodologies: IMF Balance sheet-based model and a static credit VaR using monte carlo simulation; complemented with sensitivity analysis and reverse stress testing.
  - Credit portfolio split by credit types (retail, mortgages, SMEs and commercial), currency, and main activity of corporate borrowers.
  - Stress test horizon: 3-years (2021-23).
  - Passive balance sheet assumption (i)–(iv) as specified in the source.
  - Dividend payout ratio assumed to be 50 percent in the stress tests.
- Scenario definitions:
  - Baseline: April 2021 WEO projections; economy continues recovering from April 2020 slump.
  - Adverse Scenario 1: Extended pandemic — COVID infections extend into late 2021; GDP growth remains sluggish in 2022.
  - Adverse Scenario 2: A Sudden Stop — combines Extended Pandemic with an adverse financial environment, substantial currency depreciation, and a spike in interest spreads.
- Liquidity stress test design:
  - Cash-flow cum counterbalancing capacity-based liquidity stress test using Basel III-LCR ratio.
  - Separate results by local and foreign currency.
  - Four scenarios combining two shocks to liabilities and two assumptions on asset haircuts; shock calibration based on historic data.
  - Hurdle rates: 75 percent for LCR in local currency and 100 percent for LCR in FX.
- Corporate stress test scope:
  - About 100 large corporations on a solo basis, representing about 49 corporations at the consolidated level.
  - Sample targets the ten largest borrowers of each bank.
  - Time horizon: Data unbalanced during 2018‒20.
  - Framework based on Interest Coverage Ratio (ICR) and EBITD conditional on distressed scenarios used in bank stress tests.
- Interconnectedness analysis:
  - Interbank network includes all fifteen banks.
  - Method: Balance sheet-based interbank model by Espinosa-Vega and Solé (2010).
  - Interbank positions as of October 2020, with later update to target December 2020 as final cutoff date.

*Source: NBG and IMF staff calculations and FSAP stress testing matrix as presented in the supplied content.*

### 4. Reporting

### 4. Reporting

### Format for Results / Output presentation
- Interbank network: a network chart, index of vulnerabilities.
- Common exposure: system-wide capital shortfalls.
- Evolution and direction of spillovers.

### Appendix II. Stress Test Scenarios — overview
- The main set of stress tests is based on three scenarios, a baseline and two distressed scenarios, each one spanning 2021‒23.

### Baseline Scenario
- 1. The baseline scenario rules out a second COVID wave. It assumes that the economy continues recovering from the April 2020 slump, supported by benign external financial conditions. Still, growth momentum remains weak in the near term, as social distancing measures and the pandemic drags on tourism and other contact-intensive services. The unemployment rate picks up throughout 2020, hitting household income. The economy starts to recover on firmer footing in the first half of 2021, driven by a quicker rebound in manufacturing and agriculture. Real GDP reaches pre-crisis levels by early 2022 and negative effects on long-term potential prove negligible.
- 2. Monetary conditions in this scenario are supportive. The NBG maintains a low policy rate on account of subdued inflation. Throughout, housing prices are resilient, except for some segments of commercial real estate, which undergo a mild drop in real terms. In the FX market, substantial lari depreciation in 2020 is followed by a moderate but stable appreciating trend from 2021 onwards. The current account deficit widens in 2020, driven by a halt in tourism receipts, and is financed by steady remittances and official inflows. Mirroring this trend, net FDI and portfolio inflows decline in 2020 and start to recover in the second half of 2021 helping to close the large external financing needs.
- 3. Fiscal performance is adversely affected by the pandemic due to cyclical factors and the expansion of expenditures in support of households and businesses. Public debt to GDP jumps in 2020 on account of the wider fiscal deficit, real FX depreciation and some buildup of liquidity buffers. Yet, sovereign spreads stay compressed, reflecting sound fiscal fundamentals and ample liquidity provision overseas. Public debt remains sustainable and starts to decline with gradual fiscal consolidation from 2022 onwards.
- 4. In the banking system, non-performing loans increase during 2021‒22, but without threatening systemic stability. Banks maintain good profitability and solvency ratios and credit growth remains resilient. Yet, the provision of credit to SMEs and commercial real estate is sluggish.

### Scenario 1: Extended Pandemic
- 5. This scenario assumes that COVID infections extend into late 2021 due to lags in vaccine production and distribution, population resistance, or virus mutations. Weak demand and lockdown measures weigh heavily on production processes. As a result, GDP growth remains sluggish in 2022, fueling unemployment and hitting household income. The corporate sector is also embattled. Firms operating in tourism-related services, particularly SMEs are severely hit, and many become insolvent. The effects of the pandemic on economic activity are long-lasting due to the phased withdrawal of support measures, and lower potential due to scarring and other long-lasting effects of the virus on production processes.
- 6. This scenario assumes the continuation of benign external financial conditions, as reserve currency countries maintain expansionary fiscal and unconventional monetary policies. Thus, monetary conditions remain broadly supportive. Inflation declines on account of weak demand and muted FX pass-through, allowing the NBG to keep the policy rate at historic lows. Property prices undergo a substantial drop, more severely felt on commercial real estate. In the FX market, lari depreciation continues into 2021, and bouts in FX volatility are curbed by NBG intervention. The current account deficit moderates but external financing needs remain large, with official inflows and steady worker remittances helping to close the gap.
- 7. In this environment, fiscal performance is severely affected by the drop in tax revenues and the expansion of support measures to households and businesses. Public debt balloons, resuming a downward path only after 2023. Sovereign spreads widen moderately, reflecting the effects of the crisis on public finances.
- 8. The banking system experiences substantial stress. Non-performing loans increase sharply and the need for greater loan-loss provisioning, together with higher funding costs and lost interest income dent on profitability. Additional vulnerabilities include the substantial dependence on external financing and high concentration in the banking sector. As a result, credit conditions tighten and credit growth becomes negative, leading to second round effects on real activity. The tightening of credit conditions and weakening of the real economy expose vulnerabilities, with some banks requiring fresh capital.

### Scenario 2: Capital Outflows
- 9. This scenario, more severe and less plausible, assumes a progression of the pandemic parallel that of the Extended Pandemic scenario, but incorporates an adverse external financial environment. Georgia faces a sudden stop or an episode of capital outflows that leads to substantial currency depreciation, and a spike in interest spreads, with adverse feedback effects on economic activity. This configuration tries to capture risks stemming from the large stock of external liabilities, the still high dollarization, and the sizable gross external financing needs. The materialization of this scenario is triggered by an idiosyncratic shock (i.e., political turmoil in the region or concerns on the solvency of a systemic institution), or by a global shock (i.e., increases in policy rates in industrial countries leading, or increase in risk aversion and associated turmoil in global capital markets).
- 10. In this environment, real activity is heavily affected by weak demand, the impact of the pandemic on production, and adverse feedback loops stemming from credit markets. GDP growth remains in negative territory during 2022 causing a sharp and long-lasting increase in unemployment and significant financial distress on low-income and severely indebted households. Significant lay-offs are reported in the construction and retail sectors, hurting household income. Across sectors, the most affected by the slump are construction, manufacturing, consumer goods, and hotels and restaurants. There is widespread insolvency among SMEs and firms operating in tourism-related activities. The effects of the crisis on potential output are long lasting, and real GDP fails to reach pre-crisis levels during the projection.
- 11. Monetary conditions tighten. The sudden stop induces the NBG to intervene forcefully in the FX market and to hike the policy rate. Yet, inflation remains subdued in response to global commodity price developments and a sharp contraction in domestic demand. The current account deficit closes and the trade balance is fully financed by worker remittances. Net FDI and portfolio inflows decline until the first half of 2023. The drop of external financing leads to the evaporation of liquidity, forcing banks to increase rates on domestic and FX deposits and leading to margin compression. Credit growth turns negative, and banks react to deteriorating asset quality and profitability by tightening lending standards. Property prices undergo a substantial decline, and some segments of commercial real estate such as offices and retail spaces are severely hit.
- 12. The effects of this scenario on the fiscal sector are severe, as the extension of countercyclical fiscal policy combines with weak tax revenues. Public debt to GDP increases substantially due to the combined effects of FX depreciation and the fiscal gap. Sovereign spreads widen, adding to the stress.
- 13. In the banking system, liquidity and solvency ratios deteriorate. The level of nonperforming loans increases sharply and the need for greater loan-loss provisioning, higher funding costs and lost interest income dent on profitability. Additional vulnerabilities are the dependence on external financing and high banking sector concentration.

*Source: 4. Reporting — Appendix II. Stress Test Scenarios (Georgia).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2021/english/1geoea2021005.pdf_