## Management of Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia — Executive Summary

## Source details

**Canonical URL:** [Management of Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia — Executive Summary](https://www.imf.org/-/media/files/publications/dp/2022/english/mfsrccccaea.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/dp/2022/english/mfsrccccaea.pdf.md)
- [Structured JSON version](/-/media/files/publications/dp/2022/english/mfsrccccaea.pdf.json)

---

### Crisis context and near-term outlook
- COVID-19 raises the risk of renewed financial sector pressures in the Caucasus and Central Asia (CCA) region in the period ahead.
- On average, economic growth across the region was 7 percentage points lower in 2020 than expected in 2019.
- Strong policy responses in 2020–21 have delayed the full impact of the crisis, but financial sector risks will increase once public support is phased out.
- Episodes of bank distress may have large macro-financial implications: weakened balance sheets could impair banks’ ability to lend during the recovery, and severe financial sector distress could lead to costly public interventions; past fiscal costs from government interventions exceeded 20 percent of GDP in a few cases.

### Key features of CCA banking systems
- Banks represent almost 90 percent of total financial system assets on average.
- Banking sectors are concentrated: the top five banks (ranked by assets) hold between 55 and 85 percent of system assets.
- State ownership and quasi-fiscal activities are high in three countries (Azerbaijan, Tajikistan, Uzbekistan).
- Loans constitute the largest part of banks’ assets; corporate loans represent 57 percent of the total loan portfolio on average.
- Deposits are the dominant funding source, with a high share of demand deposits.
- Net interest income constitutes over 65 percent of banks’ income on average; income from fees and commissions accounts for about 20 percent.
- Banking sectors are relatively small in several countries: total banking assets are below 50 percent of GDP in Azerbaijan, Kazakhstan, Kyrgyz Republic, and Tajikistan; Georgia and Armenia have total assets of more than 90 percent of GDP.
- Interest rate spreads (lending minus deposit rates) are 8, 10, and 24 percent in Uzbekistan, Kyrgyz Republic, and Tajikistan, respectively.
- Collateral requirements surpass 200 percent of loan amounts in Armenia, Azerbaijan, Kyrgyz Republic, and Georgia.

### Credit risk and cyclicality
- Legacy of elevated problem loans from the 2014–15 oil price shock persists in many CCA countries.
- Corporate-sector credit risk dominates in Azerbaijan, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan; household-sector credit risk is higher in Armenia.
- NPL ratios are generally highest in trade and construction sectors across the region.
- Large and externally driven credit cycles have magnified systemic risk historically; in 2019, high corporate and household credit growth and positive credit-to-GDP gaps in some countries (Armenia, Georgia, Kyrgyz Republic, Uzbekistan) may have raised underlying credit quality risks.
- Foreign currency lending to unhedged borrowers has frequently added to the build-up of credit risk.

### Currency, dollarization, and FX-related risks
- Dollarization remains elevated: about 45 percent for the region, well above the average emerging market (EM) or emerging European (EMEU) economy.
- Household deposits denominated in FX exceed 50 percent in Armenia and Azerbaijan, and are 65 percent in Georgia.
- About 70 percent of corporate loans are denominated in FX in Armenia, Georgia, and Tajikistan.
- As much as 40 percent of household loans (mainly mortgages) are denominated in FX in Georgia.
- Dollarization-related risks:
  - Indirect FX credit risks due to many FX loans being to unhedged borrowers; NPLs on FX loans are higher than in domestic currency for households and firms in Georgia, and for firms in Azerbaijan, Kyrgyz Republic, and Tajikistan.
  - Dollarization can magnify FX liquidity risk in financial systems.
- Measures that supported de-dollarization in some countries include transition to inflation targeting, higher interest rates, and macroprudential measures.

### Liquidity and funding risks
- Loan-to-deposit ratios are well above 100 percent in Armenia, Georgia, and Uzbekistan (about 120, 140, and 220 percent, respectively), creating liquidity risks if deposits are withdrawn.
- FX funding reliance is high:
  - FX funding is about 50 percent in Azerbaijan, and about 60 percent of total liabilities in Uzbekistan and Georgia.
  - Loan-to-deposit ratios in FX are almost 300 percent in Uzbekistan.
  - Nonresident FX deposits exceed 20 percent in Armenia and Georgia and are more prone to withdrawal in times of stress.
- Central banks have limited capacity to provide FX liquidity support; in early 2020, several CCA central banks (Armenia, Georgia, Kyrgyz Republic) intervened to provide FX liquidity support.

### Stress test findings and system vulnerabilities
- Stress test analysis identifies FX risks as especially significant for the region’s financial systems at this juncture.
- Under adverse macroeconomic scenarios:
  - CCA banks’ capital adequacy ratios could drop significantly but would likely remain above regulatory minimums.
  - FX-induced credit risk could severely impact bank capitalization.
  - Half of the banks in the region could become illiquid under acute FX funding stress.
  - Simultaneous realization of multiple risks would have compounded effects.
  - The largest and state-owned banks appear to be the most vulnerable.

### Quantifying financial stability risks (selected stress-test results)
- Under an adverse macroeconomic scenario:
  - NPL ratios would almost triple in Azerbaijan and Kyrgyz Republic, and increase significantly in other countries, except Armenia.
  - The average NPL ratio in the region would increase to 26.7 percent in 2021 (from 12 percent in June 2020), ranging from 11.2 percent (Uzbekistan) to 66.7 percent (Tajikistan).
  - The average capital adequacy ratio (CAR) would drop from 24.8 percent to 20 percent in 2021.
  - Eight banks in the region (ranging from 13 percent to 24 percent of assets in their country) would see their CAR ratios fall below the Basel II regulatory minimum (8 percent) in the absence of policy response.
  - Only three banks (about 0.6 percent to 7 percent of assets in their country) may face bankruptcy as their CAR could turn negative.
- FX-induced credit risk sensitivity:
  - Assuming exchange rate depreciation by 2 standard deviations, the decline in CARs is about 6 percentage points on average across the region.
  - More vulnerable countries: Armenia, Kyrgyz Republic, and Uzbekistan (declines from 5.6 to 10.5 percentage points).
  - For Azerbaijan, Kazakhstan, Tajikistan, CARs would decline by less than 4 percentage points.
  - Note: 2SD FX shock is more extreme than standard credit risk shocks (1SD); out of a 5.9 percentage point decline in CARs, indirect FX risks contributed 5.7 percentage points (footnote in source).
- Liquidity stress testing:
  - HQLA appears sufficient to absorb liquidity outflows while cash inflows decline over four months in most countries, but liquidity ratios deteriorate significantly and the number of illiquid banks surges if access to central bank liquidity support is curtailed.
  - Withdrawal assumptions: domestic demand deposits 15 percent per month and other liabilities 10 percent per month lead most systems’ liquidity ratios to drop close to zero or become negative in one case.
  - FX funding pressures scenario: 25 percent run-off on FX deposits and 35 percent on other FX liabilities leads to negative liquidity ratios after four months for most CCA countries (except Kazakhstan and Azerbaijan).
  - Without policy support:
    - Scenario I (first shock): 92 out of 147 banks (63 percent of banks) will remain liquid.
    - Scenario II (more severe FX run-off): 70 out of 147 banks (48 percent of banks) will remain liquid.
- Interest rate risk sensitivity:
  - Assuming interest rate increases of 2 percent, CARs would drop by:
    - Kazakhstan: 2.4 percentage points
    - Armenia: 11.2 percentage points
    - Kyrgyz Republic: 1.4 percentage points

### Heterogeneity of impact and bank types
- Largest banks and state-owned banks seem the most vulnerable in general.
- Within countries, post-shock CARs depend heavily on initial conditions.
- Small banks are generally better capitalized in CCA countries, although in Kazakhstan and Tajikistan they appear more vulnerable given higher NPL ratios.
- In Azerbaijan, large banks have strong capital buffers.
- Liquidity shocks tend to have the largest impact on large banks, except in Azerbaijan and Kazakhstan where large banks have high liquidity buffers initially.
- Financial soundness indicators for state-owned banks in Uzbekistan, Tajikistan, Azerbaijan, and Kyrgyz Republic are generally weaker than their private counterparts.

### Asset quality, near-term challenges, and lessons from earlier crises
- Problem loans definition includes: (1) overdue loans >90 days; (2) watch loans <90 days; (3) restructured/prolonged loans; (4) write-offs; and (5) transfer to special-purpose vehicles or other vehicles.
- Banks’ sovereign bond portfolios (share of bank assets): Kazakhstan about 17 percent, Armenia 10 percent, Kyrgyz Republic 8 percent.
- Dominance of fixed-rate longer-maturity loans implies a sudden increase in funding costs could squeeze banks’ profit margins.
- Indicators of banking soundness deteriorated somewhat in the first half of 2020 and are likely to worsen: NPLs respond with a lag; bank profitability declined since end-2019 (except in Azerbaijan and Kazakhstan); loan loss provisions have not increased; bank capital ratios have receded in some countries (Armenia, Georgia, Tajikistan, Uzbekistan); equity market valuation of some CCA banks has fallen.
- Experience from the 2014–15 oil price shock:
  - NPL ratios peaked from 7 percent in Georgia and Armenia to 54 percent in Tajikistan in that episode.
  - Interventions often entailed large public sector liabilities (examples cited: Azerbaijan 22 percent of GDP support to publicly owned largest banks; Kazakhstan ~4 percent of GDP; Tajikistan 6 percent of GDP; Uzbekistan 6.6 percent of GDP).
  - Policies that restored balance sheets while limiting fiscal costs and moral hazard included supervisory measures to strengthen capitalization, deposit insurance enhancements, FX swaps, additional private capital raising, higher capital requirements prompting mergers and acquisitions, retained earnings, and stronger NPL resolution frameworks.

### Table of selected financial soundness indicator changes (percentage point change since Dec 2019)
- Armenia (ARM): Capitalization −0.6; Liquidity 10.4; Asset Quality 0.6; Loan Loss −0.5; Profitability −1.3
- Georgia (GEO): Capitalization −0.3; Liquidity 2.6; Asset Quality 0.3; Loan Loss 6.3; Profitability 16.9
- Kyrgyz Republic (KGZ): Capitalization −1.6; Liquidity −1.7; Asset Quality 3.4; Loan Loss 2.1; Profitability −0.1
- Tajikistan (TJK): Capitalization 0.6; Liquidity 29.4; Asset Quality −10.7; Loan Loss 2.8; Profitability 3.0
- Azerbaijan (AZE): Capitalization 2.9; Liquidity −3.3; Asset Quality −2.1; Loan Loss −1.6; Profitability −0.6
- Uzbekistan (UZB): Capitalization −6.1; Liquidity 3.8; Asset Quality 3.1; Loan Loss −8.2; Profitability −7.5
- Kazakhstan (KAZ): Capitalization 2.5; Liquidity 10.2; Asset Quality −1.0; Loan Loss −1.9; Profitability 1.7

### Policy recommendations and lessons — overview
- CCA bank balance sheets were resilient in the initial phase of the COVID crisis, but risks will likely increase in the near term once public support is phased out.
- Prolonged reliance on emergency measures (borrower support, relaxed prudential requirements) may reduce transparency, undermine credit discipline and threaten future financial stability.
- It is critical to have a well-communicated strategy to unwind policy support.
- Key vulnerabilities: credit risk (especially indirect FX exposures) and liquidity risk (including FX liquidity); vulnerability to sharp exchange rate depreciations; shocks from higher leverage and rising delinquencies in corporate and household sectors could erode profitability and capitalization and affect lending capacity.

### Recommended range of reforms (objectives and key actions)
- Objectives:
  - Reduce risks of bank distress and limit costly public interventions.
  - Strengthen resilience and policy frameworks.
- Supervisory frameworks — key actions:
  - Base supervisory policy on in-depth risk diagnostics, including stress tests.
  - Reestablish loan classification and provisioning rules as soon as possible.
  - Monitor NPLs by sector, large borrowers, and restructured loans; focus on FX exposure data and borrower hedging.
  - Limit capital distributions during crises.
  - Frequently monitor bank funding and liquidity positions, especially in FX; use liquidity buffers and forward-looking reporting, including Basel III liquidity ratios where introduced (Georgia, Kazakhstan).
  - Monitor exposures to government bonds and interest rate- or FX-linked instruments.
- Macroprudential perspective:
  - Upgrade macroprudential frameworks to build resilience across credit cycles; deploy tools where appropriate to limit excesses and support de-dollarization.
  - Country examples: Armenia plans a stressed loan-to-value limit; deepen domestic public debt markets to support local currency use; in data-poor countries consider simple rules-based tools.
- Bank resolution and insolvency:
  - Strengthen bank resolution frameworks and insolvency regimes to support swift balance sheet repair and limit fiscal costs.
  - Encourage NPL management capabilities, internal plans, and operational write-off targets.
  - Postcrisis strategies often combine: recapitalization/restructuring; stronger insolvency legal frameworks; moving NPLs off balance sheet via sales or Asset Management Companies.
  - Special out-of-court restructuring can be scalable when insolvency systems are overwhelmed.
- Broader structural reforms:
  - Reduce state role in financial sector, promote competition, entrench market discipline and "risk culture".
  - Strengthen institutions: supervision, credit information, legal frameworks for collateral enforcement.
  - Financial inclusion: SME access to credit is about 7 percent of total bank lending; scaling up SME finance would support diversification and growth.
  - Diversify financing sources: capital markets and fintech development; require upgraded regulatory frameworks to limit risks to financial stability, integrity, and consumer protection.

### Country lessons and reforms (selected examples)
- Georgia and Armenia: withstood the 2014–15 shock without public support due to healthy liquidity and capital buffers; NPLs increased marginally in Georgia due to stringent lending standards and proactive restructuring; capital buffers rebuilt via additional capital and retained profits.
- Georgia policy measures included raising minimum capital requirement (to GEL50 million in 2017 from GEL12 million in 2014), introduction of LCR (2017) and NSFR (2019), deposit insurance (2017–19), Pillar-2 buffers (CICR, CPCB, GRAPE), and Pillar-3 disclosures.
- Armenia raised minimum capital requirement to AMD30 billion in effect in Jan-2017 from AMD5 billion in 2014; deposit guarantee fund existed before 2014–15 shock.
- Georgia insolvency reform (2018–2020) improved recovery rates and recovery times; new insolvency law approved in September 2020 with timely implementation critical.

### Appendix — stress-test methodology and key assumptions (high level)
- Stress tests use simplified balance sheet and income statement data for banks in 6 CCA countries: Armenia, Azerbaijan, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan.
- NPL empirical models estimated with bank-level data from FitchConnect or national authorities.
- Key modelling assumption: lending is core of banking; loan losses are the main risk to profitability and capital.
- Pass-through of exchange rate changes to inflation assumed ≈ 35 percent across countries.
- Banks’ pre-impairment net income assumed to be zero for capital impact calculations.
- Two macro scenarios: December 2020 baseline and a more adverse V-shaped scenario (decline in real GDP slightly larger than baseline by 2 SD for most countries; larger exchange rate depreciation).
- Credit risk mapping to CAR follows a logistic NPL model, mapping projected NPLs into substandard/doubtful/loss categories with provisioning rates of 20 percent, 50 percent, and 100 percent respectively, and CAR recalculated accordingly.
- FX shock assumptions include depreciations as large as 2014–15 shock and a 2 standard deviations FX shock over the last 15 years.
- Liquidity scenarios:
  - Scenario 1: domestic demand deposits run-off 15 percent per month and other liabilities 10 percent per month.
  - Scenario 2: FX run-off 25 percent on FX deposits and 35 percent on other FX liabilities.
- Interest rate repricing shocks: increases of 2 and 5 percentage points considered in Armenia, Kazakhstan, and Kyrgyz Republic.
- Caveats: top-down approach may miss bank-specific factors; some countries use aggregate data; stress-test horizon limited to 2021; models may be subject to structural breaks; interbank linkages not fully quantified; extraordinary policy measures not fully factored in.

*Management of Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia — Executive Summary (IMF staff).*

### Executive Summary ................................................................................................v

### Executive Summary

### Crisis context and near-term outlook
- The COVID-19 crisis raises the risk of renewed financial sector pressures in the Caucasus and Central Asia (CCA) region in the period ahead.
- On average, economic growth across the region was 7 percentage points lower in 2020 than expected in 2019.
- Strong policy responses in 2020–21 have delayed the full impact of the crisis, but financial sector risks will increase once public support is phased out.
- Episodes of bank distress may have large macro-financial implications: weakened balance sheets could impair banks’ ability to lend during the recovery, and severe financial sector distress could lead to costly public interventions; past fiscal costs from government interventions exceeded 20 percent of GDP in a few cases.

### Key features of CCA banking systems
- Banks represent almost 90 percent of total financial system assets on average.
- Banking sectors are concentrated: the top five banks (ranked by assets) hold between 55 and 85 percent of system assets.
- State ownership and quasi-fiscal activities are high in three countries (Azerbaijan, Tajikistan, Uzbekistan).
- Loans constitute the largest part of banks’ assets; corporate loans represent 57 percent of the total loan portfolio on average.
- Deposits are the dominant funding source, with a high share of demand deposits.
- Net interest income constitutes over 65 percent of banks’ income on average; income from fees and commissions accounts for about 20 percent.
- Banking sectors are relatively small in several countries: total banking assets are below 50 percent of GDP in Azerbaijan, Kazakhstan, Kyrgyz Republic, and Tajikistan; Georgia and Armenia have total assets of more than 90 percent of GDP.
- Interest rate spreads (lending minus deposit rates) are 8, 10, and 24 percent in Uzbekistan, Kyrgyz Republic, and Tajikistan, respectively.
- Collateral requirements surpass 200 percent of loan amounts in Armenia, Azerbaijan, Kyrgyz Republic, and Georgia.

### Credit risk and cyclicality
- A legacy of elevated problem loans from the 2014–15 oil price shock persists in many CCA countries.
- The corporate sector is a larger source of credit risk in Azerbaijan, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan; household-sector credit risk is higher in Armenia.
- NPL ratios are generally highest in trade and construction sectors across the region.
- Large and externally driven credit cycles have magnified systemic risk historically; in 2019, high corporate and household credit growth and positive credit-to-GDP gaps in some countries (Armenia, Georgia, Kyrgyz Republic, Uzbekistan) may have raised underlying credit quality risks.
- Foreign currency lending to unhedged borrowers has frequently added to the build-up of credit risk.

### Currency, dollarization, and FX-related risks
- Dollarization remains elevated: about 45 percent for the region, well above the average emerging market (EM) or emerging European (EMEU) economy.
- Household deposits denominated in FX exceed 50 percent in Armenia and Azerbaijan, and are 65 percent in Georgia.
- About 70 percent of corporate loans are denominated in FX in Armenia, Georgia, and Tajikistan.
- As much as 40 percent of household loans (mainly mortgages) are denominated in FX in Georgia.
- Dollarization generates several financial sector risks:
  - Indirect FX credit risks arise because many FX loans are to unhedged borrowers; NPLs on FX loans are higher than in domestic currency for both households and firms in Georgia, and for firms in Azerbaijan, Kyrgyz Republic, and Tajikistan.
  - Dollarization can magnify FX liquidity risk in financial systems.
- Measures that have supported de-dollarization in some countries include transition to inflation targeting, higher interest rates, and macroprudential measures.

### Liquidity and funding risks
- Loan-to-deposit ratios are well above 100 percent in Armenia, Georgia, and Uzbekistan (about 120, 140, and 220 percent, respectively), creating liquidity risks if deposits are withdrawn.
- FX funding reliance is high:
  - FX funding is about 50 percent in Azerbaijan, and about 60 percent of total liabilities in Uzbekistan and Georgia.
  - Loan-to-deposit ratios in FX are almost 300 percent in Uzbekistan.
  - Nonresident FX deposits exceed 20 percent in Armenia and Georgia and are more prone to withdrawal in times of stress.
- Central banks have limited capacity to provide FX liquidity support; in early 2020, several CCA central banks (Armenia, Georgia, Kyrgyz Republic) intervened to provide FX liquidity support.

### Stress test findings and system vulnerabilities
- Stress test analysis identifies FX risks as especially significant for the region’s financial systems at this juncture.
- Under adverse macroeconomic scenarios:
  - CCA banks’ capital adequacy ratios could drop significantly but would likely remain above regulatory minimums.
  - FX-induced credit risk could severely impact bank capitalization.
  - Half of the banks in the region could become illiquid under acute FX funding stress.
  - Simultaneous realization of multiple risks would have compounded effects.
  - The largest and state-owned banks appear to be the most vulnerable.

### Policy recommendations and lessons from past crises
- A range of policies are needed to preemptively address risks, building on international best practices and lessons from CCA experiences:
  - Supervisory frameworks: base supervisory policy on in-depth risk diagnostics; continue transitions to risk-based supervision and address weaknesses (related-party lending, corporate governance, accounting, underwriting).
  - Macroprudential perspective: upgrade macroprudential frameworks to build resilience across credit cycles; deploy macroprudential tools where appropriate to limit excesses and support de-dollarization.
  - Bank resolution and insolvency: strengthen bank resolution frameworks and insolvency regimes to support swift balance sheet repair and limit fiscal costs.
  - Broader structural reforms: reduce the role of the state and promote competition in banking systems; diversify financing sources through capital market and fintech development to support safe and sustainable credit growth and financial inclusion.
- Past experience from the 2014–15 crisis shows that appropriate policy responses—such as strengthened prudential regulation or decisive measures to rebuild capital—can help maintain market confidence and restore buffers without incurring large fiscal costs.

_Management of Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia — Executive Summary_

### 1. Asset Quality

### 1. Asset Quality

### Asset Quality and Exposure
- Problem loans definition (panel 1): (1) overdue loans >90 days; (2) watch loans <90 days; (3) restructured/prolonged loans; (4) write-offs; and, (5) transfer to special-purpose vehicles or other vehicles.
- Banks’ sovereign bond portfolios (share of bank assets): Kazakhstan about 17 percent, Armenia 10 percent, Kyrgyz Republic 8 percent.
- Given dominance of fixed-rate longer-maturity loans, a sudden increase in funding costs (for example, shock to deposit interest rates) could squeeze banks’ profit margins.
- Where a large share of government bonds is held in the available-for-sale portfolio, they are subject to market risk and marked to market.

### Quantifying Financial Stability Risks (Stress Testing)
- Under an adverse macroeconomic scenario:
  - NPL ratios would almost triple in Azerbaijan and Kyrgyz Republic, and increase significantly in other countries, except Armenia.
  - The average NPL ratio in the region would increase to 26.7 percent in 2021 (from 12 percent in June 2020), ranging from 11.2 percent (Uzbekistan) to 66.7 percent (Tajikistan).
  - The average capital adequacy ratio (CAR) would drop from 24.8 percent to 20 percent in 2021.
  - Eight banks in the region (ranging from 13 percent to 24 percent of assets in their country) would see their CAR ratios fall below the Basel II regulatory minimum (8 percent) in the absence of policy response.
  - Only three banks (about 0.6 percent to 7 percent of assets in their country) may face bankruptcy as their CAR could turn negative.
  - Note: While the risk of bank failure appears to be low, it could be higher if the quality of capital is not sufficiently loss-absorbing.
- FX-induced credit risk sensitivity:
  - Assuming exchange rate depreciation by 2 standard deviations, the decline in CARs is about 6 percentage points on average across the region.
  - More vulnerable countries: Armenia, Kyrgyz Republic, and Uzbekistan (declines from 5.6 to 10.5 percentage points).
  - For Azerbaijan, Kazakhstan, Tajikistan, CARs would decline by less than 4 percentage points.
  - Assuming exchange rate depreciations similar to the 2014–15 shock leads to comparable results.
  - Footnotes in source: 2SD FX shock is more extreme than standard credit risk shocks (1SD); out of a 5.9 percentage point decline in CARs, indirect FX risks contributed 5.7 percentage points.
- Liquidity stress testing:
  - Banking systems’ stock of high-quality liquid assets (HQLA) appears sufficient to absorb liquidity outflows while cash inflows decline over a period of four months in most countries, but liquidity ratios deteriorate significantly and the number of illiquid banks surges if access to central bank liquidity support is curtailed.
  - When the withdrawal of domestic demand deposits and other liabilities is assumed to be 15 and 10 percent per month, respectively, most banking systems would see liquidity ratios drop close to zero or become negative in one case.
  - FX funding pressures scenario: assuming run-off rates of 25 percent on FX deposits and 35 percent on other FX liabilities, liquidity ratios after four months would be negative for most CCA countries (except Kazakhstan and Azerbaijan).
  - FX deposits by domestic residents represent the largest share in total FX funding in most countries and are relatively stable funding sources (especially where a strong deposit insurance system covering FX deposits is in place).
  - Without policy support:
    - Scenario I (first shock): 92 out of 147 banks (63 percent of banks) will remain liquid.
    - Scenario II (more severe FX run-off): 70 out of 147 banks (48 percent of banks) will remain liquid.
- Interest rate risk sensitivity:
  - Assuming interest rate increases of 2 percent, CARs would drop by:
    - Kazakhstan: 2.4 percentage points
    - Armenia: 11.2 percentage points
    - Kyrgyz Republic: 1.4 percentage points
  - Most of the impact comes from repricing, with little impact through net interest income.

### Heterogeneity of Impact and Bank Types
- Largest banks and state-owned banks seem the most vulnerable in general.
- Within countries, post-shock CARs depend heavily on initial conditions.
- Small banks are generally better capitalized in CCA countries, although in Kazakhstan and Tajikistan they appear more vulnerable given higher NPL ratios.
- In Azerbaijan, large banks have strong capital buffers.
- Liquidity shocks tend to have the largest impact on large banks, except in Azerbaijan and Kazakhstan where large banks have high liquidity buffers initially.
- Financial soundness indicators for state-owned banks in Uzbekistan, Tajikistan, Azerbaijan, and Kyrgyz Republic are generally weaker than their private counterparts.

### Near-Term Challenges and Lessons from Earlier Crises
- Removal/unwinding of policy support needs careful timing and management; unwinding is likely to exacerbate vulnerabilities including exposures to oil price volatility, elevated debt levels and financing needs, weak asset quality, and high dollarization.
- Exchange rate depreciations could magnify credit risk; economies more integrated in the global financial system (Azerbaijan, Georgia, Kazakhstan) may face risks from weaker trade and capital outflows.
- Indicators of banking soundness deteriorated somewhat in the first half of 2020 and are likely to worsen going forward: NPLs respond with a lag; bank profitability declined since end-2019 (except in Azerbaijan and Kazakhstan); loan loss provisions have not increased; bank capital ratios have receded in some countries (Armenia, Georgia, Tajikistan, Uzbekistan); equity market valuation of some CCA banks has fallen.
- Experience from the 2014–15 oil price shock:
  - NPL ratios peaked from 7 percent in Georgia and Armenia to 54 percent in Tajikistan in that episode.
  - Interventions often entailed large public sector liabilities (examples cited: Azerbaijan 22 percent of GDP support to publicly owned largest banks; Kazakhstan ~4 percent of GDP; Tajikistan 6 percent of GDP; Uzbekistan 6.6 percent of GDP).
  - Policies that restored balance sheets while limiting fiscal costs and moral hazard included supervisory measures to strengthen capitalization (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic), deposit insurance enhancements (Armenia, Georgia, Tajikistan), FX swaps (Armenia, Kazakhstan), additional private capital raising, higher capital requirements prompting mergers and acquisitions, and retained earnings.
  - Stronger NPL resolution frameworks contributed to restoring bank capital buffers.
  - Appropriate policy responses supported credit recovery in several countries (credit growth increased in Armenia, Georgia, Kyrgyz Republic, Uzbekistan after the shock).

### Table 1: Selected Financial Soundness Indicator Changes (percentage point change since Dec 2019)
- Armenia (ARM): Capitalization −0.6; Liquidity 10.4; Asset Quality 0.6; Loan Loss −0.5; Profitability −1.3
- Georgia (GEO): Capitalization −0.3; Liquidity 2.6; Asset Quality 0.3; Loan Loss 6.3; Profitability 16.9
- Kyrgyz Republic (KGZ): Capitalization −1.6; Liquidity −1.7; Asset Quality 3.4; Loan Loss 2.1; Profitability −0.1
- Tajikistan (TJK): Capitalization 0.6; Liquidity 29.4; Asset Quality −10.7; Loan Loss 2.8; Profitability 3.0
- Azerbaijan (AZE): Capitalization 2.9; Liquidity −3.3; Asset Quality −2.1; Loan Loss −1.6; Profitability −0.6
- Uzbekistan (UZB): Capitalization −6.1; Liquidity 3.8; Asset Quality 3.1; Loan Loss −8.2; Profitability −7.5
- Kazakhstan (KAZ): Capitalization 2.5; Liquidity 10.2; Asset Quality −1.0; Loan Loss −1.9; Profitability 1.7

*Source: IMF staff calculations and national authorities as presented in the source PDF.*

### Box 1. (continued)

### Box 1. (continued)

### Macro-Financial Responses to Coronavirus
- Most supervisors did not reverse the relaxation of loan classification and provisioning rules, with the exception of the Kyrgyz Republic.
- Most supervisors exited from lower liquidity requirements, while maintaining lower capital requirements.
- Most countries in the region continued to restrict dividend payments in 2021.
- Authorities in the region strengthened financial safety nets and contingency planning in case of a relapse of the pandemic.
- Box Figure 1.1 lists country-level measures (easing of countercyclical capital buffers; modification of loan classification and provisioning rules; regulatory response on liquidity requirements; regulatory response on capital requirements; adjustment of other macroprudential policies; loosening of macro-financial policy response; public announcement on COVID-19-related policies affecting bank lending) with country annotations including AZE, KAZ, KGZ, ARM, GEO, TJK, UZB.

*Sources: National authorities; and IMF staff calculations.*

### Box 2. Georgia and Armenia: Policies Supporting Balance Sheet Recovery
- Georgian and Armenian banks withstood the 2014–15 oil price shock without public support due to healthy liquidity and capital buffers.
- NPLs in Georgia increased only marginally because of stringent lending standards, high profitability and proactive loan restructuring.
- Capital buffers were quickly rebuilt after the shock through additional capital and retained profits.
- NPLs gradually declined as loans were restructured and/or restarted performing, or were written off in a timely way, supported by effective insolvency regimes.
- Bank profitability (ROE) improved immediately after the shock and remained relatively high thereafter, aiding capital restoration; higher profits were driven by a combination of credit growth, loan restructuring and NPL write-offs.
- Raising banks’ minimum capital requirements and introducing additional capital buffers:
  - Georgia: minimum capital requirement raised to GEL50 million in 2017 up from the GEL12 million in 2014.
  - Armenia: minimum capital requirement increased to AMD30 billion in effect in Jan-2017 from AMD5 billion in 2014.
- Market consolidation and supervisory actions:
  - In Georgia, the two largest banks (Bank of Georgia and TBC Bank) each bought smaller banks; the license of a third bank was revoked.
  - In Armenia, higher capital requirements triggered M&As; number of commercial banks fell from 21 to 17.
- Strengthening of financial stability frameworks and tools:
  - Both NBG and CBA introduced capital buffers for systemically important banks (D-SIB), capital conservation buffer (CCB), and the countercyclical capital buffer (CCyB).
  - The NBG imposed additional Pillar-2 capital requirements such as the unhedged Currency-Induced Credit Risk (CICR) buffer, the credit portfolio concentration buffer (CPCB), and the General Risk Assessment Program (GRAPE) buffer.
  - The GRAPE guidelines integrate stress tests with Pillar-2 buffers under the Basel framework.
  - NBG introduced Basel Pillar-3 disclosure requirements and gained oversight of credit information bureaus.
  - Introduction of liquidity coverage ratios in Georgia, and deposit insurance in 2017-19, strengthened resilience against liquidity risks.
  - Armenia had established a deposit guarantee fund before the 2014–15 oil price shock; deposit insurance schemes in both Armenia and Georgia covered both local and FX deposits.
  - In 2017, the CBA formally added financial stability as a primary objective along with price stability.

### 5. Conclusions and Policy Recommendations — Overview
- CCA bank balance sheets were resilient in the initial phase of the COVID crisis, but risks will likely increase in the near term once public support is phased out.
- Stronger pre-pandemic bank buffers in the CCA than in other EMs and EMEU, along with decisive policy interventions, helped alleviate the economic and financial fallout of the pandemic in 2020.
- Prolonged reliance on emergency measures (borrower support, relaxed prudential requirements) may reduce transparency, undermine credit discipline and threaten future financial stability.
- It is critical to have a well-communicated strategy to unwind policy support.

### Key Vulnerabilities Identified
- Stress tests show key vulnerabilities relate to credit risk, especially indirect risk through FX exposures, and liquidity risk, including in foreign currency.
- CCA countries are vulnerable to sharp exchange rate depreciations.
- Shocks from higher leverage and rising delinquencies in the corporate and household sectors could erode banks’ profitability and capitalization and affect lending capacity.

### Recommended Range of Reforms (Objectives)
- Reduce risks of bank distress and limit costly public interventions.
- Strengthen resilience and policy frameworks through:
  - (i) Supervisory policies to repair balance sheets and preserve/enhance bank resilience.
  - (ii) Bank restructuring and resolution policies, and private sector insolvency regimes to limit impact of bank failures.
  - (iii) Macroprudential policy to shield against systemic shocks.
  - (iv) Broader reforms to facilitate greater financial inclusion and provision of safe and sustainable credit.

### A. Supervisory Frameworks — Key Actions
- Supervisory strategies should be guided by in-depth risk diagnostics, including stress tests.
- Require banks to maintain close monitoring and reporting on underlying asset quality; reestablish loan classification and provisioning rules as soon as possible.
- Credit risk:
  - Monitor NPLs by sector, large borrowers, and restructured loans for forward-looking capital assessment.
  - Focus on FX exposure data and borrower hedging given importance of indirect FX-related credit risk.
  - Limit capital distributions (dividend payouts, share buybacks, bonus payments) during crises.
- Liquidity risk:
  - Frequently monitor bank funding and liquidity positions, especially in foreign currency.
  - Use liquidity buffers and liquidity management tools based on forward-looking reporting—including Basel III liquidity ratios where introduced (Georgia, Kazakhstan).
- Market risk:
  - Monitor exposures to government bonds and interest rate- or FX-linked instruments; assess market illiquidity, volatility, and issuer/counterparty risk.
- Good-practice lessons:
  - Georgia: holistic reforms, Pillar-2 capital requirements, LCR (2017), NSFR (2019), frequent stress tests in 2020 informing provisioning and capital plans.
  - Kazakhstan: following a 2019 AQR, new regulatory agency introduced RBS in 2020 including SREP covering financial sustainability, business models, corporate governance, capital and liquidity risks; plan to conduct stress tests and AQRs regularly.
  - Kyrgyz Republic: NBKR developed a three-year Strategic Plan (2017–2019) to adopt RBS and aim to fully implement Basel III standards, prepare and adopt ICAPP and SREP enhancements, and ensure capital reflects banks’ risk profiles.

### B. Macroprudential Perspective
- Macroprudential policy can build resilience given CCA susceptibility to large cycles, external shocks, credit and asset price booms/busts, and structural vulnerabilities (Khandelwal and others 2021).
- Macroprudential tools can help reduce dollarization and should complement macroeconomic stability and exchange rate flexibility.
- Most CCA countries have started upgrading macroprudential frameworks; tools to build capital buffers and contain sectoral risk build-up are relatively recent; tools to mitigate FX risks have been used for more than 10 years in many CCA countries.
- Country-specific examples and actions:
  - Armenia plans to adopt a stressed loan-to-value limit to further de-dollarize.
  - Efforts to develop/deepen markets for domestic public debt would support local currency use.
  - In countries with limited data/analytical capacity (Kyrgyz Republic, Tajikistan, Turkmenistan, Uzbekistan), simple and rules-based tools can increase financial system resilience (IMF 2014).
  - In countries with better data and deeper markets, such as Georgia, more sophisticated time-varying tools can be considered.
- Footnote references:
  - Pillar 2 capital requirements cover buffers related to credit, liquidity, market, operating, business model, and profitability risks; consider macroeconomic environment, group structure, corporate governance.
  - Kazakhstan AQR: covered 14 largest banks, accounting for 87 percent total banking assets; identified capital shortfall of about $1 billion as of April 2019, concentrated in 4 banks; measures reduced this amount by half; remainder expected to be covered by shareholders with state guarantees or central bank loan in one case.

### C. Bank Resolution Frameworks and Insolvency Regimes
- Stronger bank resolution frameworks and insolvency regimes are needed to reduce scarring risks from the COVID crisis, financial stability risks, and contingent public liabilities.
- Swift and efficient NPL resolution is critical for balance sheet repair and value recovery.
- COVID likely to leave large amounts of distressed assets; banks should develop internal NPL management capabilities, plans, and ambitious operational write-off targets.
- Successful postcrisis strategies often combine three Pillars:
  1. Bank recapitalization or restructuring.
  2. Strengthening insolvency legal frameworks.
  3. Moving NPLs off balance sheet via bilateral sales or Asset Management Companies.
- Special out-of-court restructuring can provide scalable solutions when insolvency systems are overwhelmed; standardized restructuring solutions (Iceland, 2010–11) offer less-costly options for small businesses and households.

### D. Broader Structural Reforms
- Long-term safe and sustainable credit growth requires holistic reforms:
  - Reduce state role in financial sector (ownership, credit subsidies, guarantees), promote competitive banking systems, entrench market discipline and "risk culture".
  - State-owned banks generally have higher NPLs and lower profitability.
  - Strengthen institutions: sound supervision and regulation, credit information availability, strong legal frameworks to enforce collateral and lower cost of credit.
- Financial inclusion:
  - CCA countries have one of the lowest levels of SME access to credit in the world (about 7 percent of total bank lending—MCD DP 19/02).
  - Scaling up SME finance would yield macroeconomic benefits and support diversification and growth.
- Diversifying sources of financing:
  - Capital markets and fintech remain underdeveloped in CCA.
  - Armenia launched a medium-term capital market development strategy in July 2020 to develop new markets and establish regulatory frameworks.
  - Fintech adoption could increase access to finance and reduce cost, mobilize private savings, create digital cashflow and collateral trails, enhance credit risk assessment, improve AML/CFT compliance, and develop SME-tailored lending instruments.
  - Upgraded regulatory frameworks are required to limit risks to financial stability, integrity, and consumer protection (MCM DP 20/09).

### Box 3. Georgia: Resolving Insolvency
- Georgia has a strong and rapidly improving framework for resolving borrower insolvency; improved considerably from 2014 as indicated by recovery rates and World Bank Resolving Insolvency indicator.
- Recovery time: two years.
- Recovery rate: above 40 percent (against cost of 10 percent).
- Since 2018, with IMF technical assistance, Georgia has been revamping its insolvency law to align with international standards.
- Shortcomings of the 2007 Insolvency Law included: inefficient and costly processes; insufficient protection of creditors including secured creditors; lack of a proper rehabilitation framework; lack of tools for optimal debt restructuring and asset disposal; capacity and resource constraints at the National Bureau of Enforcement (NBE) which was the sole insolvency service provider and charged high fees.
- 2018 reform highlights:
  - Transform NBE into a supervisory agency; create a new class of insolvency professionals to assume managerial duties.
  - Introduce a pre-insolvency framework to facilitate debt restructuring with limited court involvement.
  - Protect all creditors’ rights via the “best interest of creditors’ test” for dissenting creditors; allow secured creditors to vote on restructuring plans affecting their rights.
  - Strengthen rehabilitation framework with remedies for breach of procedural deadlines, added flexibility in approval timeframes, and wider restructuring tools (debt-to-equity swaps, decision-making on executory contracts).
  - Liquidation/bankruptcy processes: allow credit-bidding, secured creditor sale of security under certain conditions, and explicit capping of administrative expenses associated with asset sales.
- In September 2020, Parliament approved the new insolvency law; timely implementation (including licensing insolvency professionals) is critical to handle potential increases in corporate insolvencies post-COVID-19.

*IMF DEPARTMENTAL PAPERS • Managing Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia*

### Appendix 1. Methodology and

### Appendix 1. Methodology and Assumptions for Stress Tests

### Overview of stress testing approach
- Stress tests use simplified balance sheet and income statement data for banks in 6 CCA countries: Armenia, Azerbaijan, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan.
- NPL empirical models were estimated using annual bank-level data from FitchConnect (Azerbaijan, Kazakhstan) or data from the authorities (Armenia, Kyrgyz Republic, Tajikistan, Uzbekistan).
- Key modelling assumption: lending is the core of banking; loan losses are the main risk to banks’ profitability and capital.
- Pass-through of exchange rate changes to inflation is assumed to be about 35 percent across countries.
- Banks’ pre-impairment net income is assumed to be zero for the capital impact calculations.

### Macroeconomic scenarios applied
- Two macroeconomic scenarios applied to each country: the December 2020 baseline scenario and a more adverse scenario (V-shaped recovery) that simulates a sharper recession and larger exchange rate depreciation.
- Baseline scenario characteristics:
  - Real GDP falls sharply in 2020 and rebounds strongly in 2021 (as projected in the December 2020 WEO).
  - Exchange rate depreciation is large in most CCA countries in 2020 and expected to depreciate by less in 2021.
  - Inflation driven by changes in growth and exchange rates (pass-through ≈ 35 percent).
  - Interest rates increase to mitigate inflationary and depreciation pressures, deepening the impact on NPLs.
  - Remittances fall in 2020 and are expected to rebound in 2021 in line with the recovery in Russia.
- Adverse scenario characteristics:
  - Decline in real GDP is slightly larger than in the baseline scenario (2 SD from the mean for most countries).
  - Real GDP recovers in a V-shape in 2021 similar to the baseline scenario.
  - Exchange rate depreciation is higher with the sharper recession.
  - Consideration of policy-intervention unwinding to identify pockets of vulnerabilities and hidden risks.
- U- or L-shaped recoveries were not considered systematically due to high uncertainty about potential growth-path assumptions.

### Credit risk methodology (NPL modelling and mapping to CAR)
- Three-step approach to estimate macroeconomic impact on solvency:
  1. Estimate elasticities between NPLs and macroeconomic indicators using a dynamic panel model with bank-specific fixed effects.
  2. Use elasticities to calculate implied change in NPLs consistent with the macroeconomic scenarios.
  3. Estimate loan loss provisions resulting from the change in NPLs; additional provisions are directly drawn from CAR given pre-impairment net income = zero.
- NPL model specification (logistic transformation):
  - LNPL denotes the logistic transformation of NPLs ratio: LNPL = ln(NPL/(1 − NPL)).
  - Regressors: bank fixed effects (μi), lagged NPL, real GDP growth (gt), inflation (inflt), interest rate (rt), percentage changes in bilateral exchange rates (DEPt – NEER or Spot rate in USD), percentage changes in remittances (REMITt).
  - Standard errors computed using Heteroskedasticity-consistent Robust standard errors.
- Projection formula used:
  - NPL_t^stress = [1 / (1 + ((NPL_{t-1}^{initial} / (1 − NPL_{t-1}^{initial}))^−1 * exp{b ΔX_t}))] (notation reflects the logistic-transformation-based update where X_t is vector of macro factors and β the estimated coefficients).
- Interpretation: holding other variables constant, the odds of loans becoming non-performing change by 100 × (exp{β} − 1) for a one-unit increase in an independent variable.
- Mapping projected NPLs to provisioning and CAR (Annex Figure 1.1 steps):
  1. Projected NPL (from NPL regression model, NPL).
  2.1 Substandard loan (SL) = NPL × %Substandard loan.
  2.2 Doubtful loan (DL) = NPL × %Doubtful loan.
  2.3 Loss loan (LL) = NPL × %Loss loan.
  3.1 Provision for substandard loan (PSL) = SL × Provision rate for SL (20%).
  3.2 Provision for doubtful loan (PDL) = DL × Provision rate for DL (50%).
  3.3 Provision for loss loan (PLL) = LL × Provision rate for LL (100%).
  4. Provision required = PSL + PDL + PLL.
  5. New capital = Existing capital − (Provision required − Existing provision).
  6. CAR = New capital ÷ Risk-weighted assets.

### Key empirical findings on NPL determinants
- NPLs are very persistent across CCA regions; lagged NPLs are the most statistically significant determinant with logit loadings ranging from 0.3 to 0.8.
- Real GDP growth is a statistically significant determinant of NPLs (negative correlation) in all countries; elasticities vary:
  - A lower growth of 1 percentage point raises the ratio (logarithmic scale) of NPLs by 0.015 in Uzbekistan to 0.097 in Azerbaijan.
- For CCA oil-importing (OI) countries, remittances are a statistically significant driver of NPLs:
  - Lower remittances of 1 percentage point increase the ratio (logarithmic scale) of NPLs by 0.002 to 0.01.
- Statistical significance of loadings on inflation, exchange rate, and interest rate varies across countries; some coefficients are significant, others are mixed.

### Foreign exchange, liquidity, and interest rate risk scenarios
- Foreign exchange risk assessment components:
  - Direct solvency risk from banks’ net open FX positions and local-currency positions indexed to exchange rates.
  - Indirect solvency risk from FX positions taken by borrowers affecting their ability to repay.
  - FX liquidity risk from liquidity mismatches in foreign currency.
- FX shock assumptions include depreciations as large as those in the 2014–15 oil price shock and a 2 standard deviations FX shock over the last 15 years.
- Indirect FX risk sensitivity tests assume higher elasticities of NPLs to exchange rate shocks (for example, half or more FX loans become NPLs), exceeding what the NPL model implies.
- Liquidity stress test scenarios:
  - Scenario 1: Withdrawal of domestic demand deposits and other liabilities assumed to be 15 and 10 percent per month respectively (this shock is at the lower bound of deposit-run magnitudes assumed in IMF FSAPs).
  - Scenario 2: FX liquidity stress: 25 percent run-off rates on FX deposits, and 35 percent run-off rates on other FX liabilities (consistent with historical episodes in the region).
- Interest rate risk:
  - Repricing shocks assume interest rates increase by 2 and 5 percentage points in Armenia, Kazakhstan, and Kyrgyz Republic (similar to shocks experienced in 2014–15).
  - Losses assessed on total government bond portfolios from repricing if banks sell and mark-to-market sovereign bonds.
  - Interest rate risk analysis not undertaken for other CCA countries due to lack of supervisory data.

### Caveats and limitations
- Top-down approach promotes uniformity and consistency but may obviate important country- or bank-specific factors.
- Use of aggregate data in a few countries (Armenia, Tajikistan) due to data limitations; Armenia model estimated at aggregate level, Tajikistan at aggregate level due to lack of historical bank-by-bank NPL data.
- Credit stress test horizon limited to 2021 because of forecast uncertainty; full credit losses from COVID-19 may require longer horizons.
- NPL model may be subject to structural breaks; relationships between NPLs and growth could deviate from historical patterns if large shocks or new policy measures occur.
- Stress tests did not quantify risks from domestic and cross-border interbank linkages due to data scarcity.
- Stress tests did not fully factor in extraordinary policy measures implemented during the COVID-19 crisis, which may delay or smooth some impacts on banking systems.

*Source: IMF staff calculations, Appendix 1. Methodology and Assumptions for Stress Tests, Managing Financial Sector Risks from the COVID-19 Crisis in the Caucasus and Central Asia, Departmental Paper No. DP/2022/005.*

---


_Source: https://www.imf.org/-/media/files/publications/dp/2022/english/mfsrccccaea.pdf_