## _cr09171

## Source details

**Canonical URL:** [_cr09171](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2009/_cr09171.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2009/_cr09171.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2009/_cr09171.pdf.json)

---

### Main findings and conclusions
- Market-based indicators and stress tests complement accounting-based FSIs to assess Cyprus’s financial stability and can be further developed to improve forward-looking analysis.
- Various indicators place Cyprus’s banking system soundness ahead of emerging countries but behind advanced economies; market-based indicators give a more mixed picture than accounting ratios.
- Cypriot commercial banks are weaker than counterparts in Greece and less stable than cooperative banks in Cyprus, driven by higher return volatility of commercial banks.
- Credit risk is the main source of risk; an extreme but plausible stress scenario implies costs of 4 percent GDP.
- Historical banking crises had average fiscal costs of 15–20 percent of GDP (Appendix IV).
- Spillover analysis:
  - Strong co-movement between the Cypriot banking equity index and bond spreads in Russia and Ukraine during recent turmoil.
  - Correlation with Greek bank equity index almost doubled in the past three years.
  - Asymmetric contagion: shocks to individual Greek banks significantly impact Cypriot banks’ default probability while the reverse does not hold.
- Recommendation summary:
  - Strengthen CBC financial stability capacity (add resources; improve integration with supervision).
  - Develop forward-looking market-based indicators and more robust stress-testing frameworks.
  - Encourage banks to reduce counterparty risks and maintain prudential liquidity standards.

*Prepared by Martin Čihák and Heiko Hesse, with inputs from Kalin Tintchev and Srobona Mitra; stress testing done in cooperation with Joseph Theodorou and Marianna Charalambous (Central Bank of Cyprus).*

### Banking stability: key indicators and comparisons
- Aggregate profitability and capital (commercial banks, as of end-2007):
  - Return on equity (ROE): 37 percent
  - Return on assets (ROA): 2 percent
  - Tier I capital to risk-weighted assets: 10 percent
- Cooperative credit institutions (aggregate data to end-2006):
  - Regulatory capital to risk-weighted assets (2003–2006): 11.1, 11.3, 12.1, 13.8
  - Regulatory Tier I capital to risk-weighted assets (2003–2006): 11.1, 11.3, 12.1, 13.8
  - Return on assets (2003–2007): -0.1, 0.7, 0.6, 0.7, 0.7
  - Return on equity (2003–2007): -1.9, 10.4, 8.4, 9.4, 9.4
- Balance-sheet ratios (commercial banks series):
  - Customer deposits to total (non-interbank) loans (1999–2008 March): 131.9, 133.1, 137.0, 132.0, 127.6, 140.9, 134.7, 132.2, 122.8, 109.7
  - Liquid assets to total assets (commercial banks): 29.5, 31.1, 29.6, 25.2, 26, ... (table series)
  - Net open position in foreign exchange to capital (series includes): 2.9, 7.6, 5.8, 1.9, 2.9, ... , 0.4, 1

### Key financial soundness comparisons (aggregates)
- IMF 2002–07 averages (%):
  - Capital to Risk-Weighted Assets: Cyprus 12.2; Europe 12.8; World 17.8; Advanced 15.5
  - NPLs to Total Loans: Cyprus 8.1; Europe 2.3; World 6.5; Advanced 6.9
  - Return on Assets: Cyprus 0.6; Europe 0.7; World 1.5; Advanced 1.5
  - Return on Equity: Cyprus 10.9; Europe 15.4; World 14.7; Advanced 15.6
- ECB 2006 statistics (Cyprus vs EU comparators):
  - Net interest income (% of total assets): Cyprus 2.0; EU Avg 2.5; Median 2.0; Stdev 2.4
  - Net non-interest income: Cyprus 0.9; EU Avg 1.7; Median 1.6; Stdev 0.9
  - Total expenses: Cyprus 1.4; EU Avg 2.1; Median 1.9; Stdev 1.1
  - Profits (after tax) (ROA): Cyprus 0.9; EU Avg 1.6; Median 1.1; Stdev 2.0
  - Profits (after tax) (% Tier 1) (ROE): Cyprus 14.6; EU Avg 20.3; Median 18.8; Stdev 6.9
  - Net interest income (% of total income): Cyprus 70.1; EU Avg 55.9; Median 54.4; Stdev 10.8
  - Cost-to-income ratio (% of total income): Cyprus 48.7; EU Avg 52.7; Median 53.5; Stdev 11.4
  - Overall solvency ratio: Cyprus 13.6; EU Avg 14.3; Median 12.3; Stdev 5.3
  - Tier 1 ratio: Cyprus 10.7; EU Avg 12.0; Median 10.4; Stdev 6.0
  - Liquid asset ratio (cash and loans to credit institutions): Cyprus 60; EU Avg 8.8; Median 23.2; Stdev 101.3
  - Loans to customers (% of total assets): Cyprus 53.7; EU Avg 55.7; Median 58.0; Stdev 10.4
  - Amounts owed to customers: Cyprus 80.4; EU Avg 52.3; Median 49.6; Stdev 19.2

### Z-score and volatility findings
- Z-score (1995–2007, BankScope):
  - Z-score: Cyprus 12.14; Greece 45.73; Portugal 33.1
  - ROA in %: Cyprus 0.8; Greece 1.0; Portugal 0.7
  - Equity/Assets (in %): Cyprus 6.9; Greece 7.4; Portugal 5.9
  - St. dev. of ROA (in % points): Cyprus 0.7; Greece 0.4; Portugal 0.2
- Z-score (2003–2007):
  - Z-score: Cyprus 11.74; Greece 46.9; Portugal 37.7
  - ROA in %: Cyprus 0.8; Greece 1.0; Portugal 0.9
  - Equity/Assets (in %): Cyprus 6.9; Greece 7.3; Portugal 6.5
  - St. dev. of ROA (in % points): Cyprus 0.7; Greece 0.4; Portugal 0.2
- Interpretation:
  - Lower z-score in Cyprus driven mainly by higher standard deviation of returns.
  - Cooperative banks vs commercial banks (2003–07):
    - Commercial Banks Z-score: 11.72; ROA 0.8; Capital/Assets 6.9; St. dev. of ROA 0.7
    - Cooperative Banks Z-score: 36.9; ROA 0.5; Capital/Assets 12.0; St. dev. of ROA 0.3

### Interest margins and drivers
- Net interest margins (Commercial banks):
  - 2007: Cyprus 2.59; Greece 3.52; Portugal 1.91
  - 2006: Cyprus 2.36; Greece 3.46; Portugal 1.96
  - 2005: Cyprus 2.46; Greece 3.45; Portugal 2.06
  - 1995-2007: Cyprus 2.60; Greece 3.44; Portugal 2.27
- Regression findings:
  - Overhead Costs are the dominant driver of interest rate margins.
  - Cyprus Dummy negative in several specifications, supporting lower margins in Cyprus versus Greece.

### Risks and vulnerabilities
- Credit risk and real estate:
  - Real estate prices stagnated in January–June 2008 (BuySell Index).
  - Household indebtedness: 140 percent (household debt-to-GDP) by end-December 2007 (129 percent in 2006).
  - CBC reduced maximum loan-to-value (LTV) ratio as a prudential buffer (changed to 80 percent for first-home buyers and 70 percent for vacation home).
- Counterparty and liquidity risk:
  - Cypriot banks are net lenders to the European money market with balances of 12 billion Euro with other euro area EU banks and almost 10 billion euro with the rest of the world.
  - Foreign short-term deposits (notably Russian short-term deposits) form a sizable part of foreign deposits; dependency on double-tax treaties is a related vulnerability.
  - CBC requires banks to invest at least 70 percent of foreign-currency deposits into liquid assets.
- Market and reputational risks:
  - Equity prices of main Cypriot banks declined by over 60 percent from their peak in October 2007.
  - Stock market liquidity: 2008 average daily turnover €7 Million (down from €16 million in 2007).
  - Banking sector concentration: three largest banks hold a high share of assets; assets of largest bank close to 300 percent of GDP.
  - Rating context: Fitch (2008) systemic risk/macro prudential strength C2; three largest banks rated BBB to A-; Moody’s banking strength C-; Cyprus long-term deposits A3.

### Market-based indicators and spillovers
- Market-based indicators advantages and limits:
  - Advantages: high frequency, forward-looking.
  - Limitations: poor performance when securities are thinly traded; market prices reflect losses to security holders not necessarily depositors; modeling assumptions may miss extreme events.
- Bank stock evidence:
  - Bank stocks dominate Cyprus market; almost perfect correlation between stock market index and bank stock index.
  - Cypriot bank stocks had high sensitivity to EU bank stock downturns with estimated beta about 2.5 (regression y = 2.5031x - 150.31; R^2 = 0.7659).
  - Distance-to-default (DD) measures broadly in line with other advanced economy banks, but DD weakened in 2007–08.
- GARCH spillover findings:
  - Increased co-movement between Cypriot DDs and Russian bond spreads during summer 2008 turmoil; implied correlations hit -0.20 to -0.25 at times for Russia/Ukraine versus Cyprus.
  - Models show higher co-movements with Russia/Ukraine than with Egypt/Lebanon over January 2003–July 2008.
  - No clear correlation break with U.S. liquidity proxies (U.S. Libor-OIS spread; 5-year on-the-run vs off-the-run spread).
  - Cross-country: Cyprus shows smaller spillover from Russian bond spreads than some Eastern European countries; contributing factor: 90 percent of equity ownership in Cyprus is domestic.
- Co-exceedance (contagion) results:
  - Sample: five large Cypriot banks accounting for 95 percent of system assets (1999–2007).
  - Largest spillovers originate from Greek banks; shocks in Greek banks strongly affect Cypriot bank DDs while the reverse is not evident.
  - Co-exceedances capture conditional probabilities of joint extreme negative ΔDD events (bottom 10 percent tail; threshold −0.016).

### Stress testing: methodology and results
- Approach:
  - Combined single-factor sensitivity and macroeconomic scenario analysis.
  - Two top-down exercises (CBC supervisory data; FSAP team public data) and bottom-up exercise by banks; results cross-checked.
  - Data limited to publicly available sources required simplifying assumptions.
- Sensitivity analysis highlights:
  - Credit shocks have largest potential impact; illustrative CBC test assumed NPLs double and loss-given-default equals 75 percent.
  - Exchange rate direct shocks (15 percent, 30 percent) produce negligible direct losses for most banks due to small net open positions; extreme indirect test assumed 50 percent depreciation with 5 percent of FX loans nonperforming requiring 50 percent provisioning.
  - Interest rate risk impact relatively small; most mortgage loans are flexible and linked to Euribor.
  - Liquidity: extreme deposit outflow test assumed 30 percent of demand deposits withdrawn with banks relying only on liquid assets and no lender-of-last-resort; banks’ liquidity buffers found reasonably high.
  - Counterparty risk: exposures of 12 billion euros to European banks and 9.8 billion euros to rest of world; mitigation included migrating bond investments to larger financial institutions.
- Scenario analysis (extreme combined scenario, aggregated from four large banks):
  - Assumptions: increase in NPLs by 100 percent; real estate price decline by 30 percent; decline in interest rates by 200 basis points; exchange rate depreciation by 30 percent.
  - Result: aggregate capital adequacy would decline to about 3 percent in both 2006 and 2007.
  - Fiscal/systemic implication: capital injections equivalent to about 4 percent of GDP required to ensure all banks satisfy minimum CAR of 8 percent.
- Additional conclusions:
  - Banks resilient to medium-size shocks but vulnerable under extreme shocks which could generate systemic solvency problems.
  - Stress tests for FX, interest rate, and equity price risks indicate relatively small impacts versus credit risk.

### Scenario design, calibration, and historical crisis evidence
- Crisis sample (17 capital-account crises; 1990s–2000s) empirical parameters:
  - Average exchange rate depreciation one year after crisis: 33 percent.
  - Average increase in NPL ratio a year after crisis: 9.6 percentage points; cross-country standard deviation 9.1.
  - Recovery rate sub-sample average: 59 percent; standard deviation 33 percent (implying bulk recovery rates 10–40 percent; LGD 60–90 percent).
  - Suggested baseline provisioning parameterization: 60 percent provisioning; sensitivity with 90 percent provisioning.
- Scenario design principles:
  - Scenarios should be "extreme but plausible" and reflect multiple concurrent shocks.
  - Calibration options: i) past Cyprus experience, ii) explicit model for Cyprus, iii) historical crisis episodes elsewhere, iv) scenarios from other FSAPs.
  - Limitations: scarcity of relevant historical Cypriot episodes; DSGE models of limited use for large shocks; attempts to estimate satellite models for NPLs did not yield robust results for large shocks.

### Recommendations: capacity, data, and modeling improvements
- Institutional strengthening:
  - Strengthen Financial Stability Section at CBC with additional suitably qualified graduate staff and data-collection staff to allow more analytical tasks and less data-collection burden.
  - Formalize cooperation between Financial Stability Section and Banking Supervision and Regulation Division to integrate micro- and macro-prudential work.
- Stress-testing framework improvements:
  - Include sensitivity tests on counterparty risk in top-down and bottom-up exercises.
  - Monitor banks' liquidity (assets and liabilities) with extra care and high frequency.
  - Assess how much of local banks’ balance sheets could be refinanced by the ECB under normal circumstances and with haircuts.
- Specific actionable items (timing indications):
  - Issue guidelines to banks on stress testing and monitor quarterly submissions of stress test results. (medium term)
  - Develop an off-site early warning system model to compare with stress testing results. (short term)
  - Develop contagion tests for banks, non-banks, insurance companies. (medium term)
  - Collect more comprehensive data on real estate prices; CBC to compile residential and commercial property valuation indices. (short term)
  - Conduct regular scenario analysis top-down stress tests in addition to sensitivity analysis. (medium term)
  - Work with Statistical Service to compile reliable data on household debt service to income ratios and collect disaggregated household asset/liability/income data via survey. (short/medium term)
  - Implement credit risk modeling improvements including a simplified credit risk VaR (Credit Risk+) for system-wide assessment. (medium term)

### Credit Risk modeling and Credit VaR (Credit Risk+)
- Rationale:
  - Priority to strengthen credit risk modeling because credit risk dominates stress-test losses.
  - Credit VaR produces loan loss distributions and estimates economic capital (unexpected losses) at different confidence levels (example: 99 percent credit VaR).
- Credit Risk+ features and inputs:
  - Can run bottom-up (detailed portfolio) or top-down (aggregated supervisory data preserving concentration and correlations).
  - Typical segmentation: largest five corporate borrowers (about one-third of loan book), SMEs, mortgages, consumer loans (eight sub-portfolios described).
  - Required inputs per segment: PD, size of exposure, number of loans, recovery rate; advanced version adds PD volatility and sectoral exposures.
  - Closed-form analytical solution; runs quickly on standard hardware.
- Limitations and interim approaches:
  - Cypriot banks use standardized approach under CRD (Basel II) and have not calculated historical PDs; interim proxies such as NPL ratios usable.
  - Appendix Table 16 provides example input data for illustrative Bank X and Bank Y portfolio breakdowns (PDs, Exposures, Number of Loans, Recovery Rates, Standard Deviations of PDs).
- Practical utility:
  - Useful for supervisory risk-based assessments of economic capital sufficiency.
  - Useful for financial stability scenario analysis by mapping macro shocks into stressed PDs and new loss distributions.
  - IMF-developed Credit Risk+ toolbox available to authorities and shared with Bank of Cyprus staff during stress-test work.

### Technical appendices and methodologies (summary)
- Z-score decomposition and panel regressions: dependent variable z ≡ (k + μ) / σ; regressors include bank-specific variables, industry variables, bank-type interactions, macro variables, country and year dummies.
- Margin analysis: panel model with net interest margin as dependent variable; regressors include overhead, liquidity, equity/assets, GDP growth, inflation.
- GARCH/DCC methodology: multivariate DCC by Engle (2002) estimated via three-stage procedure on daily data January 2003–end July 2008.
- Co-exceedance calculation:
  - ΔDD constructed as weekly (5 trading-day) percentage changes in DD computed daily; bottom 10 percent threshold for exceedances is −0.016.
  - Conditional binomial logit model with controls for local term structure slope, domestic stock volatility, global volatility, and European index returns; parameters α, ρ, γ interpret sensitivities to common shocks, own past extremes, and other banks’ extremes.
- Scenario design guidance (Appendix IV): multi-factor scenarios recommended, including "disorderly global adjustment" and real estate shock; calibration options and data limitations discussed.

*Source: IMF staff analysis in the provided content unit.*

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

### _cr09171 - References

### Main findings and conclusions
- Market-based indicators and stress tests complement accounting-based FSIs to assess Cyprus’s financial stability and can be further developed to improve forward-looking analysis.
- Various indicators place Cyprus’s banking system soundness ahead of emerging countries but behind advanced economies; market-based indicators give a more mixed picture than accounting ratios.
- Cypriot commercial banks are weaker than counterparts in Greece and less stable than cooperative banks in Cyprus, driven by higher return volatility of commercial banks.
- Credit risk is the main source of risk. An extreme but plausible stress scenario implies costs of 4 percent GDP. Historical banking crises had average fiscal costs of 15–20 percent of GDP (Appendix IV).
- Spillover analysis shows strong co-movement between the Cypriot banking equity index and bond spreads in Russia and Ukraine during recent turmoil, rising correlation with Greek bank equity index (correlation almost doubled in the past three years), and asymmetric contagion where shocks to individual Greek banks significantly impact Cypriot banks’ default probability while the reverse does not hold.
- CBC’s financial stability work could be strengthened by adding resources to the financial stability section and improving integration with banking supervision; additional specific recommendations are provided in the note.

### Banking stability: key indicators and comparisons
- Aggregate profitability and capital (commercial banks, as of end-2007):
  - Return on equity (ROE): 37 percent
  - Return on assets (ROA): 2 percent
  - Tier I capital to risk-weighted assets: 10 percent
- Trends and drivers:
  - Very high profitability ratios were driven in part by decreasing NPL ratios since 2004 (thresholds for classifying loans as non-performing were tightened from 9 months in 2005 to 6 months in 2005 and to 3 months in 2006).
  - Cypriot banks have lower margins than in Greece, consistent with intensifying deposit competition.
  - Cooperative credit institutions (aggregate data available to end-2006) show regulatory capital to risk-weighted assets series: 11.1, 11.3, 12.1, 13.8 (2003–2006); regulatory Tier I capital to risk-weighted assets: 11.1, 11.3, 12.1, 13.8 (2003–2006).
  - Cooperative banks: return on assets series (2003–2007): -0.1, 0.7, 0.6, 0.7, 0.7; return on equity series (2003–2007): -1.9, 10.4, 8.4, 9.4, 9.4.
- Balance-sheet structure highlights:
  - Customer deposits to total (non-interbank) loans (commercial banks): 131.9, 133.1, 137.0, 132.0, 127.6, 140.9, 134.7, 132.2, 122.8, 109.7 (1999–2008 March)
  - Liquid assets to total assets (commercial banks): 29.5, 31.1, 29.6, 25.2, 26, ... (table series)
  - Net open position in foreign exchange to capital (commercial banks series): 2.9, 7.6, 5.8, 1.9, 2.9, ... , 0.4, 1

### Risks and vulnerabilities
- Credit risk and real estate:
  - Real estate prices stagnated in January–June 2008 (BuySell Index).
  - Household indebtedness: 140 percent (household debt-to-GDP).
  - High household indebtedness raises vulnerability to a substantial decline in property prices and to higher collateral and credit risk.
  - CBC reduced the maximum loan-to-value (LTV) ratio as a prudential buffer.
- Counterparty and liquidity risk:
  - Cypriot banks are net lenders to the European money market with balances of 12 billion Euro with other euro area EU banks and almost 10 billion euro with the rest of the world (non-euro area EU banks and non-EU banks).
  - Foreign short-term deposits (notably Russian short-term deposits) form a sizable part of foreign deposits; dependency on double-tax treaties for foreign short-term deposits is a related vulnerability.
  - CBC’s conservative liquidity ratio and substantial liquidity buffers have been supportive; banks plan to reduce counterparty risks in interbank activities and bond investments.
- Market contagion and spillovers:
  - Strong co-movement between the Cypriot banking sector equity index and Russian and Ukrainian bond spreads during market turmoil.
  - Correlation between Cypriot and Greek bank equity indices has almost doubled in the past three years.
  - Shocks originating in individual Greek banks have significant impact on Cypriot banks’ default probability; reverse impact is not evident.

### Stress testing and scenario results
- Overall conclusions from stress tests:
  - Credit risk is the dominant source of potential losses; exposure has grown in recent years.
  - Banks are resilient to medium-size shocks but vulnerable under extreme shocks, which could generate solvency problems for the system.
  - An extreme but plausible scenario implies recapitalization costs of 4 percent GDP.
  - Stress tests for foreign exchange, interest rate, and equity price risks indicate relatively small impacts on banks’ capital adequacy compared with credit risk.
- Methodology and exercises referenced:
  - Bottom-up stress testing calculations were carried out by individual commercial banks based on assumptions and methodology by the FSAP team, in cooperation with CBC.
  - Appendixes provide technical background: Z-Score analysis and margin decomposition; GARCH methodology; calculating co-exceedances; designing macroeconomic scenarios; improvements in credit risk modeling.

### Policy recommendations (summary)
- Strengthen CBC financial stability capacity:
  - Add resources to the financial stability section to allow more in-depth analysis and less data-collection burden.
  - Improve cooperation and integration between CBC’s financial stability function and CBC’s banking supervisors to better link micro- and macro-prudential work.
- Enhance stress testing and market-based indicators:
  - Further develop forward-looking market-based indicators and more robust stress-testing frameworks.
  - Encourage banks to reduce counterparty risks in interbank activities and bond portfolios.
  - Maintain prudential liquidity standards and monitor exposure to foreign short-term deposits and double-tax treaty dependencies.

*Prepared by Martin Čihák and Heiko Hesse, with inputs from Kalin Tintchev and Srobona Mitra; stress testing done in cooperation with Joseph Theodorou and Marianna Charalambous (Central Bank of Cyprus).*

### 13.      There are no clear signs yet that the current global financial turmoil has directly

### 13.      There are no clear signs yet that the current global financial turmoil has directly affected banks in Cyprus, but some indirect effects are being increasingly felt.

### Impact of the global financial turmoil on Cypriot banks
- End-of-year results for 2007 are strong with profits at record levels.
- Aggregate ratio of customer deposits to (non-interbank) loans at 123 percent at end-2007.
- Banks are required to invest at least 70 percent of foreign-currency deposits into liquid assets.
- Equity prices of the main Cypriot banks have declined by over 60 percent from their peak in October 2007.
- Cypriot banks are net lenders in the European money market, but:
  - Counterparty risk has become an important issue due to bond portfolios of Cypriot banks.
  - Possible credit risk arises because most loans have flexible interest rates linked to the Euribor.
- The freeze in interbank markets has led to an increase of funding costs for European banks across the board.

### Cross-country financial sector comparison (structure and soundness)
- Financial depth and structure:
  - Ratios of private credit to GDP and liquid liabilities to GDP both exceed 100 percent of GDP in Cyprus.
  - Assets of the largest Cypriot bank stand at close to 300 percent of the country’s GDP.
  - Stock market liquidity: in 2008, on average only €7 Million was traded daily on the Cyprus Stock Exchange, down from €16 million in 2007.
- Market concentration:
  - Banking sector is relatively concentrated; three largest banks hold a high share of banking assets.
- Comparative ratings and assessments:
  - Fitch (2008) rates systemic risk and macro prudential strength of the Cypriot banking system at C2 (scale A1 to E3); same as Malta, San Marino, and Slovenia; weaker than Czech Republic, Portugal, or Greece (all B2).
  - The three largest banks have Fitch ratings BBB to A-.
  - As of November 2007, Moody’s banking strength of C- ranked average in global comparison; Cyprus long-term deposits (A3) ranked below average.

### Key Financial Soundness Indicators (2002–07 averages and ECB 2006 stats)
- IMF statistics, 2002-07 averages (%):
  - Capital to Risk-Weighted Assets: Cyprus 12.2; Europe 12.8; World 17.8; Advanced 15.5
  - NPLs to Total Loans: Cyprus 8.1; Europe 2.3; World 6.5; Advanced 6.9
  - Return on Assets: Cyprus 0.6; Europe 0.7; World 1.5; Advanced 1.5
  - Return on Equity: Cyprus 10.9; Europe 15.4; World 14.7; Advanced 15.6
- ECB statistics, 2006 (Cyprus vs EU Avg/Median/Stdev where given):
  - Net interest income (% of total assets): Cyprus 2.0; EU Avg 2.5; Median 2.0; Stdev 2.4
  - Net non-interest income: Cyprus 0.9; EU Avg 1.7; Median 1.6; Stdev 0.9
  - Total expenses: Cyprus 1.4; EU Avg 2.1; Median 1.9; Stdev 1.1
  - Profits (after tax) (ROA): Cyprus 0.9; EU Avg 1.6; Median 1.1; Stdev 2.0
  - Profits (after tax) (% Tier 1) (ROE): Cyprus 14.6; EU Avg 20.3; Median 18.8; Stdev 6.9
  - Net interest income (% of total income): Cyprus 70.1; EU Avg 55.9; Median 54.4; Stdev 10.8
  - Cost-to-income ratio (% of total income): Cyprus 48.7; EU Avg 52.7; Median 53.5; Stdev 11.4
  - Overall solvency ratio: Cyprus 13.6; EU Avg 14.3; Median 12.3; Stdev 5.3
  - Tier 1 ratio: Cyprus 10.7; EU Avg 12.0; Median 10.4; Stdev 6.0
  - Liquid asset ratio (cash and loans to credit institutions): Cyprus 60; EU Avg 8.8; Median 23.2; Stdev 101.3 (liquidity metric shown with notable cross-country dispersion)
  - Loans to customers (% of total assets): Cyprus 53.7; EU Avg 55.7; Median 58.0; Stdev 10.4
  - Amounts owed to customers: Cyprus 80.4; EU Avg 52.3; Median 49.6; Stdev 19.2

### Z-score analysis — comparative stability of commercial banks
- Z-score decomposition (BankScope data), 1995–2007 (Cyprus vs Greece vs Portugal):
  - Z-score: Cyprus 12.14; Greece 45.73; Portugal 33.1
  - ROA in %: Cyprus 0.8; Greece 1.0; Portugal 0.7
  - Equity/Assets (in %): Cyprus 6.9; Greece 7.4; Portugal 5.9
  - St. dev. of ROA (in % points): Cyprus 0.7; Greece 0.4; Portugal 0.2
- Z-score decomposition, 2003–2007:
  - Z-score: Cyprus 11.74; Greece 46.9; Portugal 37.7
  - ROA in %: Cyprus 0.8; Greece 1.0; Portugal 0.9
  - Equity/Assets (in %): Cyprus 6.9; Greece 7.3; Portugal 6.5
  - St. dev. of ROA (in % points): Cyprus 0.7; Greece 0.4; Portugal 0.2
- Interpretation:
  - Lower z-score for Cypriot banks is driven by higher standard deviation of returns.
  - During 1995–2007, ROA and capitalization in Cyprus are below Greece but marginally above Portugal.
  - 2003–07 shows little change; Cypriot banks remain substantially weaker than Greek and Portuguese counterparts due to higher return volatility.
- Regression evidence (Table 5, Z-Score regressions, 1995–2007):
  - Robust estimation (1) and Random Effects (2) specifications; observations 199.
  - Significant coefficients (Robust, p-values in parentheses):
    - Assets (-1): 0.141 (0.001)***
    - Loans/Assets (-1): -20.831 (0.000)***
    - Liquid Assets/Deposits (-1): -1.143 (0.002)***
    - Income Diversity (-1): 4.931 (0.032)**
    - Cyprus Dummy: -8.535 (0.000)***
    - Portugal Dummy: 6.227 (0.005)***
  - Random Effects results differ; Cyprus Dummy in (2): -13.254 (0.074)*
  - R-squared (Robust): 0.423
  - Note: * significant at 10%; ** significant at 5%; *** significant at 1%.
- Cooperative vs commercial banks (2003–07):
  - BankScope Data:
    - Commercial Banks Z-score: 11.72; ROA 0.8; Capital/Assets 6.9; St. dev. of ROA 0.7
    - Cooperative Banks Z-score: 36.9; ROA 0.5; Capital/Assets 12.0; St. dev. of ROA 0.3 (FSI Data also shows cooperative Z-score 36.9 with Capital/Assets 12.4)
  - Cooperative banks appear more stable on aggregate due to substantially lower return volatility and different business objectives (maximization of consumer surplus), though substantial heterogeneity exists across individual cooperative banks.

### Interest margins and drivers
- Net interest margin comparisons (Commercial banks):
  - 2007: Cyprus 2.59; Greece 3.52; Portugal 1.91
  - 2006: Cyprus 2.36; Greece 3.46; Portugal 1.96
  - 2005: Cyprus 2.46; Greece 3.45; Portugal 2.06
  - 1995-2007: Cyprus 2.60; Greece 3.44; Portugal 2.27
- Findings:
  - Net interest margins are about 1 percent higher in Greece than Cyprus.
  - Intensifying competition for deposits in Cyprus, including recent entry by international banks, has put pressure on deposit rates and margins.
- Regression results for margins (Table 8, 1995–2007, observations 218):
  - Overhead Costs strongly positive and significant across specifications (e.g., coefficient 51.372, p-value (0.000)*** in (1)).
  - Liquidity coefficients mixed; capitalization coefficients not robustly significant.
  - Cyprus Dummy negative in several specifications (e.g., -0.312 (0.060)* in (1)), supporting lower margins in Cyprus versus Greece.
  - R-squared up to 0.498 in some specifications.
  - Conclusion: Overhead costs are the dominant driver of interest rate margins.

### Market-based indicators and their role
- Market-based indicators discussed: distance to default (DD), bond prices, credit default swaps.
- Advantages:
  - High frequency, forward-looking, widely available to independent analysts.
- Limitations:
  - Poor performance if securities are not publicly traded or trading is limited.
  - Market prices reflect losses to security holders, which may differ from losses to depositors.
  - Some measures assume asset value dynamics (e.g., lognormal) that may not capture extreme events.
- Empirical evidence:
  - Market-based indicators can outperform traditional measures when forecasting distress in individual financial institutions and predicting supervisory ratings, bond spreads, and rating agencies’ downgrades.
  - Literature finds market-based indicators improve models based on banking ratios.

*Source: IMF staff analysis in the provided content unit.*

### 22. Bank stock prices in Cyprus are likely to contain useful information. The stock

### 22. Bank stock prices in Cyprus are likely to contain useful information

### Liquidity, trading patterns, and informational content
- Stock market turnover in Cyprus is relatively low; low liquidity may create “noise” for less traded stocks.
- Trading in bank stocks is rather active and dominates the market, illustrated by the almost perfect correlation between the stock market index and bank stock index in Cyprus (Figure 2).
- Bank stocks are actively traded abroad, on the Athens Stock Exchange, and prices at the two exchanges are almost perfectly correlated.
- The correlation of the bank and stock index is based on the dynamic conditional correlation GARCH model.

### Comparative volatility and sensitivity
- Cypriot bank stocks grew much faster than an average of EU banks between late 2004 and October 2007.
- In October 2007, about two months after the EU bank stocks started falling, the Cypriot bank stocks began to fall as well, at a much higher rate than the average EU bank stocks.
- At present, both the Cypriot bank stocks and the EU bank stocks are some the 30 percent below the 2004–08 average.
- An analysis of co-movement suggests a relatively high “beta,” about 2.5, i.e., Cypriot bank stocks are relatively sensitive to a general downturn affecting bank stocks.
- Regression details presented: y = 2.5031x - 150.31; R^2 = 0.7659; sample average indices: EU banking stock index (sample average=100); Cyprus banking stock index (sample average=100).

### Market-based vs accounting indicators of banking soundness
- Market-based indicator: Distance-to-default (DD) measures for Cypriot banks are broadly in line with those for other advanced economy banks.
- Accounting indicators show improvement since 2003:
  - Loan impairment coverage strengthened to more than 75 percent of impaired loans in 2007.
  - Banks' core capital to total assets ratio at end-2007: 7.3 percent (2006: 7.1 percent).
  - At end-2007, loans represented 73 percent of deposits on aggregate.
  - Since 2005 some banks issued senior long-term debt under Euro medium-term note (EMTN) programs.
- Ratings agencies considered prospects for main Cypriot commercial banks substantially positive for 2008 but noted the challenge of maintaining income generation given changed economic conditions.

### Market indicators deterioration in 2007–08
- Market indicators suggest that bank soundness, after substantial improvement since 2003, deteriorated in 2007 and 2008.
- Stock market indicators have dramatically weakened with the latest turbulence in equity markets.
- Somewhat lower DD values in recent periods may reflect market worries that continued rapid loan growth could put capital ratios under pressure if banks cannot maintain high levels of internal capital generation.
- No CDS data on the Cypriot banks; results derived from bond prices are similar.

### Spillover analysis: GARCH model findings
- Model setup and intuition:
  - Various dynamic conditional correlation GARCH models relate DD measures of four banks (CBC, Marfin, Hellenic, Universal) to Russian bond spreads to test sensitivity of banks with higher Russian exposure.
- Key GARCH findings:
  - Co-movements between Cypriot banks’ DDs and the Russian bond spread increased during the onset of financial turbulences in summer 2008, though overall correlation magnitudes are not very large.
  - Implied correlations highest for Marfin and the Russian bond spreads, followed by CBC and Hellenic—consistent with Marfin’s large engagement in Russia.
  - In a model relating overall Cypriot DDs to bond spreads in Russia, Ukraine and Egypt, co-movements with Russia and Ukraine are higher than with Egypt with correlation magnitudes hitting -0.20 to -0.25 at times during the financial crisis.
  - A GARCH model of spillovers from the U.S. does not show any clear correlation break or increase between Cypriot stock markets and proxies for U.S. funding liquidity (U.S. Libor-OIS spread) and market liquidity (5-year on-the-run versus off-the-run spread).
    - This supports that funding liquidity capacity of Cypriot banks was not sizably affected by problems in wholesale funding market segments.
    - Cypriot banks mainly rely on retail funding and, as net liquidity providers to the money markets, have partially benefited from elevated interbanking rates.
    - As cautionary measures, a few banks issued covered bonds to secure longer-term funding.
- Equity-index co-movements:
  - An equity index of the Cypriot banking sector exhibits increased co-movement with bond spreads in Russia and Ukraine during recent financial market turbulences.
  - Correlations between the Cypriot banking index and Egypt and Lebanon bond spreads increased significantly in early 2008: correlation between Cyprus and Egypt dropped from zero to -0.30 and to -0.20 for Lebanon within two months.
  - Overall, co-movements tend to be higher for Russia/Ukraine than for Egypt/Lebanon over the sample from January 2003.

### Cross-country spillover comparisons
- Cypriot stock market exhibits higher co-movement with European and U.K. stock markets than with the U.S. market (S&P 500).
- Bank indices in Cyprus and Greece exhibit higher correlation, exceeding the correlation of Cyprus with the European index and Portuguese banks.
- GARCH analysis suggests the impact of Russian bond spread spillovers on Cyprus has been relatively smaller than on several Eastern European countries:
  - The Cypriot bank index does not have as strong a co-movement with the Russian bond spread as other countries, indicated by a lower negative correlation magnitude.
  - Some Eastern European countries (Poland, Czech Republic, Hungary) show more significant co-movements with Russian bond spreads than Cyprus.
  - Contributing factors: 90 percent of equity ownership in Cyprus is domestic; Russian investors do not hold sizable stakes in domestic banks or companies. Cyprus’s role as an international business center may have isolated it somewhat from the subprime fallout.

### Summary of GARCH findings
- Cyprus is not completely immune to spillovers from the financial crisis fallout.
- The Cypriot stock market moves with markets in Greece and Europe and shows some negative co-movement with Russian spreads during times of stress, albeit smaller than for other Eastern European countries.
- Distance-to-default measures for the Cypriot banking sector exhibit some spillover potential with bond spreads in Russia and Ukraine.
- Direct repercussions from U.S. illiquidity spirals in the form of impaired funding and market liquidity were not found for Cypriot banks.

### Deposit stability and reputation effects
- Past experience suggests a lack of deposit withdrawal when problems abroad occur, though this may not hold in the future.
- Cyprus benefited from a reputation as a “safe heaven”; past observations show little evidence that stock markets in Russia and Ukraine move in tandem with deposit balances.
- The relationship between financial conditions and deposit withdrawals may be highly nonlinear and can be highlighted by a crisis.
- The Central Bank of Cyprus’s conservative liquidity ratio has served the banks well.

### Spillover risks among individual banks: co-exceedance analysis
- Methodology and sample:
  - Data sample: five large Cypriot banks accounting for 95 percent of total Cyprus banking system assets; sample period 1999–2007.
  - Binomial logit model applied to DD data to examine the likelihood that a sizeable negative idiosyncratic shock (bottom 10 percent tail of five trading-day changes in DDs) in one large Cypriot bank would be followed by a similar shock in another large Cypriot bank (“co-exceedance”).
  - Four control variables included: changes in slope of local term structure; domestic stock price return volatility; MSCI All-Country World Index return volatility; European Index returns.
  - Objective: identify potential risk concentrations and distinguish common versus idiosyncratic shock impacts.
- Results:
  - Spillover risks are unevenly distributed among large Cypriot banks.
  - The largest spillovers occur from Greek banks: Cypriot bank DDs respond strongly to large movements in DDs for Greek banks.
  - Cypriot banks are exposed to some banks from other EU countries, but to a much lesser degree.
  - Shocks from Cypriot banks do not materially impact Greek banks, suggesting potential vulnerabilities for Cypriot banks if Greek banking problems arise.
  - High equity co-movements between Cypriot and Greek banks indicate the equity market could act as a financial transmission channel; confidence losses and rumors could also produce sudden spillovers even if Cypriot fundamentals remain sound.

*Source: IMF staff calculations and analysis as presented in the chapter text.*

### 37. A set of basic stress tests was performed to assess the resilience of the banking

### 37. A set of basic stress tests was performed to assess the resilience of the banking sector to a variety of potential shocks

### Overview of stress testing approach
- Purpose: examine potential effects on banks’ financial condition of changes in risk factors corresponding to a range of adverse events.
- Nature of shocks: can be considered extreme but with a positive, albeit small, probability of occurrence.
- Data and assumptions: tests are based solely on publicly available data, requiring simplifying assumptions; past data may capture only partly the type of extreme events that might happen in the future. Results should be treated as approximations and rely on experience from other countries, expert judgment, and sensitivity analysis.
- Methodology combined:
  - Single-factor sensitivity calculations and macroeconomic scenario analysis.
  - Top-down and bottom-up approaches: two top-down exercises (one by CBC staff using supervisory data; one by the FSAP team using publicly available financial data) and a bottom-up exercise by individual banks’ risk managers.
  - Cross-checking: results of top-down and bottom-up exercises were cross-checked; discrepancies discussed among FSAP team, authorities, and bank representatives.

### Sensitivity analysis — key findings
- Credit risk:
  - Credit shocks had a large and more widespread potential impact on banks than any other single-factor shock.
  - The impact of credit risk has declined, but only marginally.
  - One illustrative test (CBC calculation) assumed NPLs double and loss-given-default equals 75 percent.
- Real estate and household exposure:
  - BuySell index: sharply rose until January 2008, after which stagnation set in (Index, January 2004=100).
  - Household indebtedness: Cyprus household indebtedness stood at 140 percent to GDP by end-December 2007 compared to 129 percent in 2006; European average ratio was 60 percent in 2006.
  - Comparative figures (2006): Ireland, Portugal and Spain between 80-90 percent; Greece around 50 percent; Netherlands and Denmark over 120 percent.
  - Policy change: maximum LTV ratio changed to 80 percent for first-home buyers and 70 percent for vacation home.
  - Implication: high household indebtedness significantly exposes Cyprus to adverse real estate shocks; substantial real estate price declines would reduce household net worth and collateral value, increasing banks’ credit and collateral risk.
- Market risks:
  - Exchange rate risk sensitivity tests evaluated direct effects of 15 percent and 30 percent depreciation and appreciation of the euro; direct losses negligible for most banks due to very small net open positions.
  - Indirect exchange rate risk test illustrated an extreme assumption: exchange rate depreciates by 50 percent against all other currencies, 5 percent of foreign exchange loans become nonperforming, and require 50 percent provisioning.
  - Interest rate risk: impact relatively small; generally, an increase in interest rates improves banks’ situation as assets reprice more quickly than liabilities (note: most mortgage loans are flexible and linked to Euribor).
  - Equity exposure: banks’ exposure to stock prices is very small following reductions after early-2000s stock price bubble.
- Liquidity risk:
  - Deposit outflow test: extreme scenario assuming 30 percent of all demand deposits withdrawn, banks can use only liquid assets, and have no recourse to other banks or lender of last resort.
  - Result: liquidity buffers reasonably high; banks would be able to withstand substantial liquidity drain before requiring emergency liquidity support.
  - Caveat: sensitivity tests looked only at liability-side shocks; during stress even “safe” assets can see liquidity evaporate, magnifying deposit outflow impacts.
- Counterparty risk:
  - Since summer 2007, interbank markets strained; counterparty risk has been a primary reason for reluctance to lend.
  - Main Cypriot banks: limited exposure to asset-backed securities; reliance on retail deposits; net lenders to European interbank market, lending overnight and at short maturities.
  - Exposures: CBC data show 12 billion euros invested in European banks and 9.8 billion euros in rest of the world.
  - Mitigation actions: lending mainly to larger financial institutions deemed safe; plans to migrate matured bond investments to larger financial institutions to reduce counterparty risk.

### Scenario analysis — design and outcomes
- Rationale: single-factor tests approximate partial impacts; scenario analysis captures interplay of multiple risk channels amid macroeconomic stress and rising macro-financial linkages.
- Scenario inputs: scenarios informed by Cyprus experience and international episodes; relationships that worked in the past may break down during dislocation.
- Extreme combined scenario (aggregated from four large banks):
  - Assumptions: increase in NPLs by 100 percent; real estate price decline by 30 percent; decline in interest rates by 200 basis points; exchange rate depreciation by 30 percent.
  - Result: aggregate capital adequacy would decline to about 3 percent in both 2006 and 2007.
  - Fiscal/systemic implication: capital injections equivalent to about 4 percent of GDP would be needed to ensure all banks satisfy minimum CAR of 8 percent.
- Additional noted result: banks’ exposures remained broadly unchanged between 2006 and 2007; combination of shocks could substantially impact capitalization.

### Recommendations for further work on financial stability (policy and capacity)
- Institutional strengthening:
  - Strengthen Financial Stability Section at CBC with additional suitably qualified graduate staff and additional data-collection staff to enable more analytical tasks (stress testing and financial stability policy work).
  - Formalize cooperation between Financial Stability Section and Banking Supervision and Regulation Division to better integrate micro- and macro-prudential work.
- Stress-testing framework improvements:
  - Include sensitivity tests on counterparty risk in both top-down and bottom-up stress testing exercises; monitor banks' liquidity (assets and liabilities) with extra care and high frequency.
  - Conduct an exercise to assess how much of local banks’ balance sheets could be refinanced by the ECB under normal circumstances and with appropriate haircuts — i.e., the degree of comfort under an extreme scenario that Cypriot banks could become net borrowers instead of net lenders to the European money market.
- Specific actionable recommendations (timing indicated where provided):
  - Issue specific guidelines to banks on stress testing, with a view to monitoring stress test results in major banks on a systematic basis. (medium term)
    - Note: banks informed via circular at beginning of bottom-up exercise that stress tests will be quarterly and results submitted to CBC; assumptions and shocks to be reviewed.
  - Develop an off-site early warning system model for banks that could be compared with stress testing results. (short term)
  - Develop tests for financial contagion of banks, non-banks, insurance companies. (medium term)
  - Collect more comprehensive data on real estate prices. (short term)
    - CBC is compiling a residential property valuation index to be revised and enhanced in the short-term; a commercial property valuation index will also be compiled soon.
    - Statistical Service of the Republic is in final stages of compiling a residential property price index based on Eurostat-approved methodology.
  - Conduct regular scenario analysis top-down stress tests, in addition to existing sensitivity analysis. (medium term)
  - Work with Statistical Service to compile reliable data on household debt service to income ratios. (short term)
    - For financial stability, collect disaggregated data on household assets, liabilities and income through a household survey. (medium term)
  - Consider implementing suggestions for improvements in credit risk modeling detailed in Appendix V, including a simplified credit risk VaR to better assess credit risk at the system-wide level. (medium term)

### Conclusions — principal messages
- Market-based indicators:
  - Provide additional dimension to financial stability analysis; indicators show a mixed picture relative to accounting ratios.
  - Cross-country comparison places Cyprus’s banking system close to the average of advanced and emerging countries.
  - Spillover risks are not evenly distributed among banks; an equity index of the Cypriot banking sector exhibited strong co-movement with bond spreads in Russia and Ukraine during recent turbulences.
- Overall resilience:
  - Basic stress tests suggest resilience of the Cypriot banking system has improved in recent years.
  - Capital injections needed to bring all banks into compliance with minimum capital requirements have been declining.
  - Improvements reflect increasing buffers (capitalization and profitability) and generally decreasing stock market exposures.
- Risk priorities:
  - Stress tests confirm credit risk as the main source of risk.
  - Direct interest rate risk limited (loan book largely floating rate); direct exchange rate risk limited (low FX positions); equity risk moderate.
  - Liquidity: banks are liquid at present, confirmed by positive results from sudden withdrawal tests.

*Source: Excerpt from IMF FSAP chapter on stress testing and financial stability (CBC/Cyprus), pages 30–38 of provided content.*

### 56. We suggests further strengthening of the Financial Stability Section at the CBC.

### _cr09171 - 56. We suggests further strengthening of the Financial Stability Section at the CBC.

### Strengthening the Financial Stability Section at the CBC
- Recommendation: "We suggests further strengthening of the Financial Stability Section at the CBC."
- Rationale: The Section "carries out very important work."
- Strengthening entails:
  - Allocating more resources "so that it can carry out in-depth analysis of the financial sector data."
  - "Putting on a firmer footing its cooperation with bank supervision."

### Appendix I: Z-Score and Margin Analysis — Z-Score Analysis
- Primary dependent variable: the z-score as a measure of individual bank risk.
- Definition: z ≡ (k + μ) / σ, where
  - k = equity capital as percent of assets,
  - μ = average after-tax return as percent on assets,
  - σ = standard deviation of the after-tax return on assets (proxy for return volatility).
- Interpretation: "A higher z-score corresponds to a lower upper bound of insolvency risk—a higher z-score therefore implies a lower probability of insolvency risk."
- Empirical approach:
  - Decompose observed differences in z-scores into capitalization, returns, and volatility.
  - Estimate panel models of the form (notation preserved):  
    tjittjjtjtjisstjsssstjtjitji
    DCMBTITTIBz
    ,,1,1,,1,1,1,,,,
    επλπφφδγβα+++++++++=
    ∑∑∑∑∑
    −−−−−
  - Dependent variable: tji z , , for bank i in country j at time t.
  - Regressors include:
    - 1,,−tji B : vector of bank-specific variables.
    - 1−jt I : time-varying banking industry-specific variables in country j.
    - s T, 1,−tjs IT and 1,,−tjis BT : bank type and interactions between type and industry/bank-specific variables.
    - tj M , , j C, and t D : macroeconomic variables, country dummies, and yearly dummies.
- Country effects: country dummy variables (example: Cyprus dummy = 1 if country is Cyprus, 0 otherwise; Portugal is another dummy). "If banks in Cyprus are relatively weaker than banks in the benchmark country Greece, the sign of the Cyprus dummy variable would be negative."
- Bank-level controls: asset size in billions of U.S. dollars, loans over assets, cost-income ratio, and an income diversity measure (degree of diversification from net interest income).
- Country-level controls: GDP growth, inflation, and Herfindahl index (sum of squared market shares in terms of total assets).

### Appendix I: Margin Analysis
- Objective: Investigate drivers of margins in Cyprus, Greece and Portugal.
- Model: Simple panel model (Beck and Hesse, 2008):
  tittiti
  MBinM
  ,,, 
  argεγβα+++=
- Dependent variable: net interest margin as a percentage of earning assets.
- Bank-specific regressors: overhead (overhead costs / total assets), liquidity (liquid assets / deposits), equity/assets ratio.
- Macroeconomic regressors: GDP growth and inflation.
- Year effects: yearly dummies included except when macroeconomic variables are incorporated.

### Appendix II: GARCH Methodology
- Framework: multivariate GARCH with Dynamic Conditional Correlation (DCC) specification by Engle (2002) (generalization of CCC by Bollerslev (1990)).
- Purpose: account for heteroskedasticity and interpret conditional variance as time-varying risk; analyze co-movement of markets.
- Estimation: three-stage DCC procedure:
  1. Fit univariate GARCH models for each variable (five variables in the specification), yielding standard deviations it h .
  2. Obtain intercept parameters from transformed asset returns.
  3. Estimate coefficients governing dynamics of conditional correlations.
- Key matrices/notations:
  - r t : n x 1 vector of asset returns with mean zero and time-varying covariance.
  - R t : time-dependent correlation matrix.
  - D t : diagonal matrix of standard deviations implied by univariate GARCH models.
  - S : unconditional correlation matrix of residuals ε t .
  - Q t : covariance matrix; ι is a vector of ones; A and B are square, symmetric; D is the Hadamard product.
  - λ i : weight parameter with contributions of 2 1− t D declining over time.
  - κ i : parameter associated with squared lagged asset returns.
- Data: "daily data from January 2003 until the end of July, 2008"; variables are first-differenced.

### Appendix III: Calculating Co-Exceedances
- Step 1 — Distance-to-Default (DD):
  - DD measure based on Black and Scholes (1973) and Merton (1974); exposition detailed in Čihák and Ong (2007).
  - DDs across banks show common trends, implying exposure to common shocks.
- Step 2 — ΔDD construction:
  - Percentage change in DD denoted "ΔDD".
  - Calculate weekly (5 trading-day) ΔDDs on a daily basis to smooth noise and capture prolonged extreme events.
  - Formula preserved: || 5 5 − − − =Δ it itit it DD DDDD DD.     (1)
- Thresholds and exceedances:
  - Rank all it DD Δ observations across banks and calculate threshold 10 T for the bottom 10 percent tail (defined as "exceedances" or "extreme values").
  - "The threshold for the 10 th percentile left tail is calculated at −0.016."
  - Exceedances for bank i at time t are binary it y given by: 1 = it y if 10 T DD it < Δ, and 0 otherwise.     (2)
  - "The 10 percent tail is a value commonly used in the literature."
- Co-exceedance definition:
  - "A 'co-exceedance' is defined as the probability that a particular bank will experience a large negative shock as a result of shock to another bank in the sample, after controlling for common shocks."
  - Co-exceedances capture all potential spillover channels without specifying explicit links.
- Conditional probability model for co-exceedances (logistic CDF):
  - Preserved equation notation:  
    ∑∑
    +
    ∑∑
    ==
    =
    −−
    =
    =
    −−
    =
    =
    ++
    ++
    B
    j
    jtjsit
    s
    siiti
    B
    j
    jtjsit
    s
    siiti
    CCF
    CCF
    it
    e
    e
    xy
    1
    1
    5
    1
    1
    1
    5
    1
    1
    ),|1Pr(
    γρα
    γρα
    β,
  - Interpretations of parameters:
    - α: sensitivity of bank i to "common shocks" it F (real and financial developments in its own country and European/global markets).
    - ρ: sensitivity of bank i to its own extreme shocks in previous periods up to s lags ( sit C − ).
    - γ: sensitivity of bank i to extreme shocks experienced by other banks in the sample in the previous period ( 1− jt C , ij ≠ ), i.e., "co-exceedance."
  - All C variables are lagged by one period to capture interbank impacts.
  - Goodness of fit: McFadden R 2 .

- Common shocks (it F) composition:
  - it F = ( yc Δ , σ C , σ E , σ W ), denoting developments in local real economy and volatilities at country, European, and global levels.

- Country-specific market shocks (σ C ):
  - Weekly (5 trading-day) returns on country-specific stock index: weekly log-difference in local currency.
  - Volatility approximated by conditional variance from GARCH(1,1):
    - t t c X ε +=     (5)
    - 2 1 2 1 2 − − + += ttt w β σ α ε σ     (6)
  - t X is weekly local currency return; 2 t σ is GARCH volatility.
  - ARCH effect captured by lagged squared residual 2 1− t ε.
  - Lagrange multiplier tests show significant ARCH(1) effects for all stock market returns used.

- Local real economy developments ( yc Δ ):
  - Use weekly (5 trading-day) changes in term structure spreads: difference between 10-year government bond yield and 1-year government bond yield.
  - Change defined as: || 5 5 − − − =Δ t tt t yc yc yc ,     (7) where t yc is the term structure spread at time t.

- Regional market shocks (σ E ):
  - Use MSCI ACEI index (regional European free-float-adjusted market capitalization index; contains 16 developed market country indices as at June 2006).
  - Denominated in currency of country where dependent bank is located; estimate GARCH(1,1) volatility.

- Global market shocks (σ W ):
  - Use MSCI ACWI; denominated in the currency of the country where the dependent bank is located; estimate GARCH(1,1) volatility.

- Data note: overlapping weekly ΔDDs adjusted for serial correlation (lag structure) and non-synchronicity concerns noted as limited because most banks trade largely in same time zone.

### Appendix IV: Designing Macroeconomic Scenarios
- Stress testing principle: scenarios should be "extreme but plausible."
- Need: complement single-factor sensitivity analysis with multi-factor scenarios because "in real episodes of stress, multiple shocks tend to take place at the same."
- Suggested multi-factor scenarios:
  - "A (continuing) disorderly global adjustment":
    - Sharp correction of global imbalances, decline in global economic growth, significant depreciation of the U.S. dollar, abrupt decline in equity prices.
    - Euro-wide monetary policy response implies reduction in short-term interest rates.
    - Reference to FSAP stress test for Portugal (IMF, 2007) as analogous.
  - Real estate shock: "A decline in real estate prices is a plausible scenario."
- Specific Cyprus considerations: both scenarios "should take into account the open nature of the Cypriot banking system, and the substantial share of business with nonresidents."
- Other scenarios (noted as interesting but problematic to implement):
  - Contagion from other countries in the region: exposure due to parent companies of foreign-owned Cypriot banks; better assessed in a regional exercise; partially reflected in liquidity tests for large funding withdrawals by foreign banks.
  - Shocks to tourism: relevant but data limitations (no separate loan category "tourism").
  - Cyclical asynchrony: possible but not timely; FSAP for Portugal used such a scenario (Portugal failing to follow euro area recovery with steep rise in short-term interest rates and fiscal consolidation per the Stability and Growth Pact); found to have smaller impacts than "disorderly global adjustment" in Portugal.
- Calibrating scenarios — options:
  - (i) past experience from Cyprus,
  - (ii) estimating an explicit model for Cyprus,
  - (iii) using data from actual banking crises in other countries,
  - (iv) using scenarios from other FSAP missions and stress testers.
- Past experience limitations:
  - Only period of negative GDP growth in modern Cypriot history was in mid-1980s; usefulness limited due to structural changes and data availability (comparable NPL data available only since early 2000s).
  - Stock price bubble bust in early 2000s was stressful but "the impact on GDP has been relatively modest."
- Estimating a model limitations:
  - DSGE models useful for small movements; of "relatively little use for large shocks" because they revert rapidly to trend.
  - Attempts to estimate a "satellite model" mapping macroeconomic scenarios into credit risk (NPLs or loan loss provisions versus macro and bank-specific variables) "did not yield robust results" for large shocks despite various specifications including vector autoregression and vector error correction models.

*Source: _cr09171 - 56. We suggests further strengthening of the Financial Stability Section at the CBC.*

### 25.      Given the limited usefulness of past data for Cyprus for modeling how major

### _cr09171 - 25.      Given the limited usefulness of past data for Cyprus for modeling how major

### Calibration approaches and benchmarking
- Stress tests should be tailored to country-specific circumstances; past FSAPs show wide variation in types, coverage, and sizes of shocks.
- Two alternative calibration approaches when domestic past data are limited:
  - Use shocks and parameterizations applied in previous FSAPs for advanced European economies (Cihak (2007) overview; updated Tables 11–15).
  - Analyze historical crisis episodes in other countries and parameterize stress scenarios based on observed extreme events.

### Empirical evidence from 17 capital-account crises (1990s–2000s) — sample and general characteristics
- Sample: 17 crises including Indonesia, Korea (1997), Philippines, Thailand (June 1997), Argentina (December 2001), Brazil (December 1998), Mexico (December 1994), Uruguay (June 2002), Bulgaria, Czech Republic (May 1997), Russia (August 1998), Slovak Republic (August 1998), Turkey, Finland (October 1991), Sweden (November 1992), Italy, Spain, and United Kingdom.
- Comparative notes for Cyprus:
  - Cyprus would be slightly above the middle of the sample in level of economic development.
  - Cyprus is characterized by a much higher degree of economic openness than the rest of the sample.
  - Compared to the emerging market average, Cyprus has better governance and transparency standards and a higher credit to GDP ratio.

### Key empirical parameterizations derived from the crisis sample
- Exchange rate:
  - The average depreciation of the exchange rate in the sample a year after the crisis was 33 percent.
  - Many countries experienced considerably larger depreciations.
  - For Cyprus (as part of the euro area), the text notes it is unclear how to translate, but indicates that "it is not implausible to assume a 30 percent depreciation of the currency" (already assumed for sensitivity calculations).
- Credit quality (NPLs):
  - The average increase in the NPL ratio in the crisis countries was 9.6 percentage points a year after the crisis.
  - Cross-country standard deviation of the increase was 9.1 (substantial variation).
  - Note: analyzing average percentage increase in nominal NPLs would give a smaller absolute increase for Cyprus given low starting NPLs; using performing loans or total loans is preferred as proxy for exposure.
- Recovery rates and loss given default (LGD):
  - For a sub-sample of 10 systemic banking crises with data, the average recovery rate was 59 percent, with a standard deviation of 33 percent.
  - The bulk of observations lie in the 10–40 percent recovery rate range, implying a 60–90 percent loss given default range.
  - Suggested baseline provisioning parameterization: 60 percent provisioning rate as a starting parameterization, and to test sensitivity with a 90 percent provisioning rate.
- Overall impacts on banks:
  - Credit risk (including indirect interest rate and exchange rate risk) is likely to dominate overall impacts when combining credit, interest rate, and exchange rate shocks.

### Proposed downside scenarios and scenario design principles
- Downside scenarios are informed by recent Cypriot experience but draw judgmentally on the broad international crisis experience because relationships observed in normal times may break down during dislocation.
- Scenario calibration draws on:
  - Historical large exchange rate movements (examples include 20%–50% devaluations; specific FSAP examples list 30% devaluation and 40% depreciation/appreciation of Euro/Dollar).
  - Interest rate shocks seen in FSAPs (examples include 100 basis point shocks, 300 basis point increases, and larger ad hoc shocks; some FSAPs used +500, +200, +0 basis point patterns across maturities).
  - Credit shock magnitudes reflected in the 9.6 percentage point average NPL increase and the recovery/LGD experiences above.

### Recommendations for improvements in credit risk modeling (methodology and data)
- Priority: strengthen credit risk modeling because it is the most important source of risk identified by the FSAP and the part where methodology can be most improved.
- Data improvements:
  - Extend backward existing NPL data broken down by credit segment — desirable but acknowledged as practically impossible in Cyprus.
  - Use bank-level estimates on NPLs (potentially better than aggregate CBC estimates); bank-level data are noisier but increase degrees of freedom and preliminary calculations show stronger estimates.
- Analytical recommendations:
  - CBC could analyze the link between credit growth and individual bank weakness via bank-by-bank regressions of financial soundness (e.g., post-shock CAR in downside scenarios) on credit growth; such results are useful for both micro- and macroprudential purposes.
  - For corporate credit, assemble basic data on largest bank corporate borrowers and their financial statements, breakdowns of GDP and exports by sector, breakdown of bank loans by currency and borrower, average interest rates on foreign and domestic debt, debt maturities, and use of foreign-currency derivatives. EBITDA histories for companies are useful but not necessary.
  - Make credit risk (default probabilities) conditional on projected behavior of explanatory economic variables. Start with a first phase to estimate links and complete in a second phase. Given data limitations, use techniques for data-constrained environments (Segoviano, 2006; Gasha and Morales, 2004; Chan-Lau and Santos, 2006) or cross-country estimates if warranted.
  - Depending on data availability, use panel data on main sectors or aggregate data; feed impact on loan losses into individual bank balance sheets and compare with banks’ profits or capital.

### Advanced modeling: Credit Risk VaR and Credit Risk + adaptation
- Proposal: model a "credit risk VaR" to compute loan loss distributions on banks’ portfolios under baseline and stressed scenarios; this highlights credit risk concentration.
- Recommended methodology: Credit Risk + technology (adapted IMF/MCM version):
  - Provides a coherent framework to assess and stress test risks in a bank’s credit portfolio.
  - Computes economic "capital at risk" (unexpected losses) by generating the portfolio loss distribution and estimating economic capital at different confidence levels (example: 99 percent credit VaR).
  - Can run in bottom-up mode (detailed portfolio information) or top-down mode (aggregated supervisory data preserving concentration and correlations).
  - Typical segmentation proposed for a commercial bank: largest five corporate borrowers (about one-third of loan book), small and medium enterprises, mortgage loans, consumer loans (eight main sub-portfolios described).
  - Required inputs per segment: PD, size of exposure, number of loans, and recovery rate. Advanced version adds stochastic PDs (standard deviation of PDs) and sectoral exposures.
  - Limitation for Cyprus: Cypriot banks have implemented the standardized approach for credit risk under the CRD (Basel II) and have not yet calculated historical PDs; in the interim NPL ratios can be used as proxies.
  - The model is analytically closed-form (no simulations required) and runs quickly on standard hardware.
- Practical utility:
  - Useful for supervisory risk-based assessments of sufficiency of economic capital relative to actual bank capital (inputs similar to Basel II).
  - Useful for financial stability analysis to assess bank solvency under stressed macro scenarios.
  - The model has been applied in Greece and Turkey FSAPs; interest from institutions including the ECB and the Bank of Portugal.

### Implementation steps and phased workplan (implicit in recommendations)
- Phase 1: Begin statistical analysis linking macroeconomic/explanatory variables to loan losses/default probabilities using available data and data-constrained methods.
- Phase 2: Complete conditional default probability models, augment with bank-level NPL analyses, and implement Credit Risk + to generate loan loss distributions and credit VaR.
- Complementary actions: collect/improve bank-level and corporate borrower data as specified to enhance bottom-up modeling over time.

*Source: IMF FSAP chapter text (appendices and tables cited within the content).*

### Appendix Table 16. Example of Input Data for the Credit VaR Model

### Appendix Table 16. Example of Input Data for the Credit VaR Model

### Standard Presentation: “Bank X” Loan Portfolio
- Table header fields implicit: PD, Exposure (local currency), Number of Loans, Recovery Rate, Standard Deviation of the PDs, Sectoral exposures (subsequent columns).
- Corporate 1: PD 0.10, Exposure 40,71,766, Number of Loans 23, Recovery Rate 0.40, Standard Deviation of the PDs 0.02, remaining sectoral cells: 0.00, 0.30, 0.70, 0.00
- Corporate 2: PD 0.02, Exposure 71,688, Number of Loans 5, Recovery Rate 0.40, Standard Deviation of the PDs 0.01, remaining sectoral cells: 1.00, 0.00, 0.00, 0.00
- Corporate 3: PD 0.03, Exposure 603,455, Number of Loans 659, Recovery Rate 0.30, Standard Deviation of the PDs 0.018, remaining sectoral cells: 0.10, 0.00, 0.10, 0.80
- Corporate 4: PD 0.04, Exposure 001,420, Number of Loans 9, Recovery Rate 0.50, Standard Deviation of the PDs 0.02, remaining sectoral cells: 0.00, 1.00, 0.00, 0.00
- Corporate 5: PD 0.02, Exposure 372,082, Number of Loans 233, Recovery Rate 0.60, Standard Deviation of the PDs 0.012, remaining sectoral cells: 0.00, 0.00, 0.00, 1.00
- SMEs: PD 0.09, Exposure 76,563, Number of Loans 25,876, Recovery Rate 0.30, Standard Deviation of the PDs 0.045, remaining sectoral cells: 0.35, 0.25, 0.15, 0.25
- Mortgages: PD 0.01, Exposure 869,881, Number of Loans 56,872, Recovery Rate 0.75, Standard Deviation of the PDs 0.009, remaining sectoral cells: 0.35, 0.25, 0.15, 0.25
- Consumer lending: PD 0.10, Exposure 783,908, Number of Loans 195,400, Recovery Rate 0.25, Standard Deviation of the PDs 0.054, remaining sectoral cells: 0.35, 0.25, 0.15, 0.25

(Note: the table labels multiple corporate rows; numeric values preserved exactly as presented.)

### Alternative Presentation: “Bank Y” Credit Portfolio — Sectoral Exposures (in percent) 1/
- Table columns indicated: Probability of Default (PD), Loss given default (LGD), Value, Number of accounts (obligors), sector breakdown columns such as Agriculture, Trade, Etc., Total.
- Suggested portfolio lines (labels shown in source):
  - Corporate 1/        
  - Total loans to obligors of the public sector
  - o/w five largest public sector customers
  - Total loans to obligors of the private sector
  - o/w five largest private sector customers
  - Loans to SMEs
  - o/w five largest SME customers
  - Mortgage
  - Consumer lending
- Instructional note: "*Please provide a breakdown by economic sectors as well, if available.*"

### Box 1. Credit Value‑at‑Risk — Summary of the IMF description
- Purpose and tool:
  - The IMF toolbox for Credit Risk+ model can be used to estimate credit VaR and determine economic capital needed to meet unexpected credit losses.
  - Originally developed by Credit Suisse Financial Products; model is a market benchmark because of simplicity, straightforward and limited assumptions, and limited data requirements.
  - IMF staff developed a software toolbox that implements the Credit Risk+ model as part of Excel (see Avesani and others, 2006, for methodology). The toolbox is available for use by country authorities and has been shared with the Bank of Cyprus staff working on stress tests.
- Advantages relative to simple stress testing:
  - Credit VaR allows analysis of how dynamics in probabilities of default of individual borrowers could affect bank sustainability even before loans are classified as nonperforming.
  - More informative than simple spreadsheet stress testing and sensitivity analysis, which focus on increases in NPLs and provisioning.
- Key modeling assumptions (as stated):
  - An obligor characterized by a given default probability could be either in default or still performing at the end of a certain time period (e.g., one month, one quarter, one year, etc).
  - Default probabilities in each time period are constant.
  - The number of defaults that occur in any given period is independent of the number of defaults that occur in any other period.
  - Each obligor’s default probabilities are assumed to be small in order to obtain a closed form solution for the loss curve and to integrate analysis with default probabilities uncertainty and correlations.
- Main uses from supervisory and macroprudential perspectives:
  - To build the loan portfolio loss distribution for monitoring credit risk at the level of each financial institution.
  - Suitable for macroeconomic scenario analysis once integrated with a model for the PDs, where PD is linked to macroeconomic indicators.
  - Example application: stress testing resilience of an institution’s credit portfolio to an exchange rate shock by mapping the shock to stressed PDs, building the new loss distribution, and computing the new level of economic capital.

*Source: IMF staff.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2009/_cr09171.pdf_