## Romania Financial System Stability Assessment — Selected Findings (cr18163)

## Source details

**Canonical URL:** [Romania Financial System Stability Assessment — Selected Findings (cr18163)](https://www.imf.org/-/media/files/publications/cr/2018/cr18163.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2018/cr18163.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2018/cr18163.pdf.json)

---

### Executive summary and system structure
- Banking sector concentration and size
  - 36 banks account for around 80 percent of financial sector total assets.
  - The largest five banks concentrate 60 percent of total deposits and 57 percent of total loans.
  - Bank assets were at just 52.7 percent of GDP as of September 2017.
- Deposit and lending dynamics
  - Private sector deposits increased from about 48 percent of banks’ total liabilities in 2011 to about 64 percent in 2017.
  - Loan-to-deposit ratio fell from 122 percent in 2008 to 75 percent in 2017.
- FX and asset quality
  - Share of FX denominated loans in June 2017: 43 percent for corporate loans and 38 percent for household loans.
  - NPLs declined to 6.4 percent as of December 2017 from their peak of 21.5 percent following the crisis.
  - Average banking sector return on assets has grown to 1.5 percent.
- Nonbank financial lenders (NBFLs)
  - NBFLs assets represent about 11 percent of total banking sector assets as of end 2016.

### Emerging vulnerabilities
- Sovereign exposure and market risk
  - Sovereign debt exposure rose from less than 5 percent in 2008 to about 22 percent of banks’ total assets in 2016.
  - Long duration of domestic sovereign debt relative to funding duration exposes banks to interest rate risk.
- Real estate and household risks
  - Housing loans increased from 21 percent of loans to households to more than 54 percent between 2008 and 2017.
  - Prima Casa program allows for LTV ratios of up to 95 percent.
  - Majority of mortgage contracts are at variable rates.
- FX and NBFL risks
  - Large stock of FX loans (declining share but still significant) poses credit risk and potential liquidity needs in foreign currency.
  - Rapid growth of credit by NBFLs, combined with relatively thin capital levels and riskier lending practices.

### Stress testing scope, scenarios, and macro assumptions
- Scope and approaches
  - Solvency: Top-Down (TD) scenario-based stress test covering the top 12 banks (80 percent of banking sector assets).
  - Liquidity: LCR, NSFR, and cashflow-based analyses in major currencies.
  - NBFLs: single-factor sensitivity tests on capital to credit losses, exchange rate risk, and increase in funding costs for 35 firms in the special register.
- Adverse scenario characterization
  - Adverse scenario produces a V-shaped recession with a cumulative decline of GDP equal to two standard deviations relative to the baseline and a cumulative peak to trough GDP contraction of 8.1 percent.
  - Severity of the adverse scenario: 2.1 standard deviations (defined as deviation of GDP growth from the baseline in terms of historical growth volatility).
- Adverse scenario deviations from baseline (Table 2, in percentage points)
  - Real GDP: -6.5 / -14.6 / -17.8 (2018 / 2019 / 2020)
  - Policy interest rate: 2.5 / 1.0 / 0.0
  - Short-term money market spread: 0.5 / 0.0 / 0.0
  - Long-term government bond yield: 4.2 / 4.0 / 3.6
  - Real Effective exchange rate appreciation(-)/depreciation(+): 14.7 / 7.3 / 7.5
  - Nominal exchange rate appreciation(-)/depreciation(+): 7.5 / 1.6 / -0.8
  - Inflation rate (CPI): 1.3 / -1.4 / -4.1
  - Unemployment rate: 1.2 / 1.6 / 1.7
  - Memo: Baseline Real GDP growth (in percent): 4.3 / 3.4 / 2.7; Real GDP growth (in percent) under adverse: -2.4 / -5.2 / -0.6; Cumulative real GDP growth (from 2017): -2.4 / -7.5 / -8.1

### Solvency stress test methodology and key assumptions
- Coverage and treatment
  - Coverage: 12 largest banks, including nine OSIIs, accounting for 80 percent of banking sector assets.
  - Risks: credit risk (all exposures), market risk (mark to market securities, exchange rate, other mark to market positions), interest rate risk in the banking book.
- Balance sheet growth and RWAs
  - Quasi-static balance sheet with balance sheets growing in line with nominal GDP path; rate of change floored at zero percent (no deleveraging in adverse scenario).
  - Under the baseline, growth of assets calibrated to half of nominal GDP growth.
  - RWAs: standardized approach for most banks; two banks partially use IRB (86 and 67 percent of their credit RWAs).
- Behavioral and capital actions
  - Banks assumed not to change maturity profile; no new equity issuance; dividends paid at 10 percent of current period net income after taxes by profit-making banks in compliance with supervisory capital requirements.
- Hurdle rates (Table 4, in percent)
  - Baseline (OSII / N-OSII): Total Capital 11.5 / 10.5; Tier 1 Capital 9.5 / 8.5; CET1 8.0 / 7.0; Basel III regulatory minimum 4.5 / 4.5; Capital conservation buffer 2.5 / 2.5; OSII buffer 1.0 / 0.0
  - Adverse (OSII / N-OSII): Total Capital 9.0 / 8.0; Tier 1 Capital 7.0 / 6.0; CET1 5.5 / 4.5; Basel III regulatory minimum 4.5 / 4.5; Capital conservation buffer 0.0 / 0.0; OSII buffer 1.0 / 0.0

### Solvency stress test results — aggregate outcomes and drivers
- Aggregate system outcomes (12 banks)
  - Before stress: Aggregate CET1 Ratio 17.1; Leverage Ratio 9.3.
  - Baseline (end-2020): Aggregate CET1 Ratio 14.1; Leverage Ratio 7.7.
  - Adverse (end-2020): Aggregate CET1 Ratio 8.2; Leverage Ratio 4.8.
  - Asset share undercapitalized (Adverse, CET1<Hurdle*): 30.3 percent.
  - Maximum Capital Shortfall (% of GDP) under Adverse: 0.3.
- Main contributors to capital depletion in the adverse scenario
  - Credit risk: negative contribution of 750 bps in CET1 over the horizon; credit loss impairments increase from -2.2 percent of RWA in 2016 to -3.5 percent of RWA in 2019 (peak provisioning).
    - Composition of credit losses: real estate mortgages account for 54 percent of total; SME backed by real estate account for 20 percent of total.
  - Trading losses on sovereign securities portfolios: 300 bps drop in capital over the horizon (trading gains of 0.6 percent of RWA in 2017 turn to a market loss of 3.6 percent in 2018).
  - Exchange rate depreciation: important role in driving credit losses on FX loans.
  - Net interest income (NII): decreases by almost 40 percent throughout the test horizon; contribution of NII to CET1 is, on average, 280 bps per year throughout the horizon, compared to 525 bps in 2017.
- Sensitivity analyses and single-factor tests
  - A 500 bp increase in sovereign yields reduces the system Tier I capital by 510 bps; impact can be up to 970 bps for certain banks.
  - A 500 bp increase in interest rates reduces the system Tier 1 ratio by 120 bps.
  - Simultaneous default of five largest nonfinancial corporate exposures: most (though not all) banks can withstand this concentration shock.
  - A 50 percent haircut on sovereign exposures would cause a majority of banks to fall under minimum regulatory capital requirements.

### Market risk, interest-rate risk, and FX exposures
- Securities portfolios and duration
  - Debt securities portfolio among the 12 major banks: around RON 71 billion or 22 percent of total assets; 97 percent are domestic bonds.
  - Average duration of debt securities portfolio: 3.5 years.
- Interest-rate risk in the banking book
  - Romanian banks have significant positive interest rate risk; repricing gap contributes to losses in interest income as the lending yield curve flattens.
  - Lending rates are constrained as they are tied to the Lombard rate, limiting pass-through of funding cost increases.
- FX exposures and loan composition (section 6)
  - Foreign currency denominated loans reached 63 percent of total loans in 2011.
  - As of December 2016, shares in the stock were: 44 percent for corporate loans in foreign currency; 42 percent for household loans in foreign currency.
  - Share of FX in total new bank lending to the nonfinancial private sector was just below 20 percent in June 2017.
  - NBFL lending to the corporate sector is predominantly in FX: 84 percent.
  - Loans in foreign currency continue to have a higher non-performing ratio than loans in lei (aggregate comparison referenced).

### Liquidity stress test results and funding structure
- Funding composition (aggregate for 12 banks)
  - Retail funding: 52 percent; Non-operational deposits: 23 percent; Other unsecured funding: 18 percent; Secured funding: almost non-existent.
  - 56 percent of lei ASF is met by retail deposits; 60 percent of euro ASF is met by retail deposits.
- HQLA composition
  - HQLA dominated by government securities in both lei and euro.
  - Central government exposures: 78 percent in RON; 65 percent in EUR.
  - Central bank assets: 8 percent in RON; 19 percent in EUR.
- Maturity mismatch
  - 65 percent of available funding has maturity less than six months; only a quarter of funding has maturity greater than 12 months.
  - 80 percent of required stable funding is of maturity greater than 12 months.
- LCR-based results (Table 6)
  - System-wide LCRs:
    - RON (Basel III): 233%
    - RON (Retail Scenario): 184%
    - RON (Wholesale Scenario): 176%
    - EUR (Basel III): 203%
    - EUR (Retail Scenario): 167%
    - EUR (Wholesale Scenario): 159%
  - Liquidity shortfalls (RON mil):
    - RON Basel III: 785
    - RON Retail Scenario: 1,925
    - RON Wholesale Scenario: 2,384
    - EUR Basel III: 362
    - EUR Retail Scenario: 613
    - EUR Wholesale Scenario: 816
  - Share of assets (average) corresponding to the shortfalls:
    - RON Basel III: 1%
    - RON Retail Scenario: 2%
    - RON Wholesale Scenario: 3%
    - EUR Basel III: 2%
    - EUR Retail Scenario: 2%
    - EUR Wholesale Scenario: 2%
- NSFR and cashflow results
  - Almost all banks meet NSFR > 1 in euro and lei; one small bank fails NSFR on a euro basis.
  - Cashflow-based analysis: with the exception of one small bank, all banks could confront persistent and sizable withdrawals without resorting to central bank support (assuming NBR standing facility acceptance of government securities at zero haircut).

### Nonbank Financial Lenders (NBFLs) — structure, stress tests, and vulnerabilities
- Sector structure and concentrations
  - Analysis limited to 35 firms in the special register; 22 are leasing firms representing about two-thirds of total assets in the sector.
  - NBFL assets: RON 26.6 billion in September 2017, around 10.3 percent of total loans to households and nonfinancial corporations.
- Leasing companies (selected metrics)
  - Liabilities: Loans from banks represent 76 percent of total liabilities.
  - Assets: Fixed assets around 66 percent; credit to customers 22%; claims against financial institutions 5%.
  - Credit to customers: 96 percent to SMEs; close to 90 percent in foreign currency; only 4 percent to individuals.
  - Capital: 11 percent; NPLs: 13 percent.
- Non-leasing companies (selected metrics)
  - Liabilities: Loans from banks represent 87 percent of total liabilities.
  - Assets: Credit to customers 63%; claims against financial institutions 18%; fixed assets 1%.
  - Credit to customers: 67 percent to individuals (95 percent denominated in domestic currency).
  - Capital: 37 percent; NPLs: 22 percent.
- NBFL sensitivity test results (single-factor)
  - A 10 percent depreciation of the domestic currency:
    - Leasing companies: aggregate capital level decreases from 11.5 percent to -1.2 percent; several firms accounting for 82 percent of the subsector’s assets fall under the 7 percent regulatory minimum capital.
    - Non-leasing companies: capital decreases to 34 percent from 37 percent; only 1 firm falls below regulatory minimum.
  - A five percent increase in NPLs:
    - Leasing companies: capital ratio falls by 430 bps to an aggregate of 7.2 percent; several firms accounting for 72 percent of total assets fall below minimum capital requirement.
    - Non-leasing companies: impact is 330 bps, bringing sectoral capital down to 33 percent; no firm falls below regulatory minimum.
  - A 300 bps increase in the cost of funding:
    - Leasing firms: capital to assets ratio reduced by 200 bps; several institutions representing around 50 percent of the sub-sector by asset share fall below minimum capital requirement.
    - Non-leasing firms: impact of 170 bps; no firm falls below minimum.

### Interconnectedness and contagion analysis
- Domestic interbank contagion (top 12 banks + rest)
  - Contagion calibration: 100 percent loss of claims in default; 65 percent rollover of funding; 50 percent haircut on assets subject to liquidation.
  - Index of contagion (outward): ranges from 0.01 to 1.6 percent.
  - Index of vulnerability (inward): ranges from 0.07 to 1.1 percent.
  - Conclusion: limited risk of contagion through interbank exposures.
- Bank–NBFL contagion (11 banks, 34 NBFLs)
  - Individual NBFLs can have greater systemic impact than individual banks due to larger exposures to NBFLs as a share of capital.
  - Index of contagion for NBFLs (outward) ranges between 0.2 and 1.2.
  - Sector-level (two-entity) results:
    - Scenario 1 (credit shock): index of contagion for NBFLs = 8.6; index for banks = 1.7.
    - Scenario 2 (credit + funding shock): index of contagion for banks = 50 (NBFLs would lose about 50 percent of their total capital if the banking system failed).
- Direct and reputational channels to parent banks
  - Six banks own fully or partially an “own-name” NBFL with capital participations ranging from 50 to 100 percent.
  - Reputational calibration: retail deposit runoff of 20 percent; 50 percent replaceable with other funding for foreign banks; 30 percent replaceable for domestic banks; remaining financing needs met by asset sales with 50 percent haircut.
  - Quantified impact on parents: capital ratios decline by 3 to 9 percentage points; liquidity ratios drop by more than half for one bank.
  - Finding: reputational risks can be more significant than direct exposures, though high initial capital and liquidity imply banks would remain resilient.

### Cross-border exposures and vulnerabilities
- Cross-border structure
  - Foreign banks represent the majority of Romania’s banking system (27 out of 37 banks).
  - Romania’s banking system claims against banks abroad: about 4 percent of total Romanian banking sector assets.
  - Liabilities to foreign banks: about 15 percent of total assets.
  - Austrian banking sector is the top foreign counterpart on both claims and liabilities.
- Cross-border contagion scenarios and results
  - Scenario (i) funding shock only: 30 percent roll-over of funding; 50 percent haircut on assets subject to liquidation.
    - Index of contagion ≈ 0.01 percent; Index of vulnerability = 10 percent.
  - Scenario (ii) combined credit and funding shocks: 100 percent loss on claims, same rollover and haircut.
    - Index of contagion = 0.08 percent; Index of vulnerability = 15 percent.
  - Conclusion: Romanian banking distress would have little impact on other sectors on average, but Romania is relatively more vulnerable to external shocks via foreign funding.

### PD estimation and LGD approaches
- PD estimation methodology
  - Default rates estimated using a Bayesian Model Averaging (BMA) approach for four portfolios: consumer, retail mortgage, SME non-RE, SME RE using quarterly data Q1 2006–Q4 2016.
  - Large nonfinancial corporate PDs set at historically stressed levels (80th percentile) due to low-default portfolio issues.
- Selected PD estimation coefficients (Appendix Table 1)
  - Consumer: PD (t-4) 0.433 (p-value 0.000); D_GDP -0.075 (p-value 0.000); Observations 468; R-squared 0.3675.
  - Retail Mortgage: PD (t-4) 0.423 (p-value 0.000); D_GDP -0.108 (p-value 0.000); EXR (t-2) 0.037 (p-value 0.040); Observations 468; R-squared 0.5134.
  - SME nonRE: PD (t-4) 0.378 (p-value 0.000); D_GDP -0.062 (p-value 0.003); Observations 468; R-squared 0.3629.
  - SME RE: PD (t-4) 0.589 (p-value 0.000); D_GDP -0.076 (p-value 0.000); EXR (t-2) 0.046 (p-value 0.008); Government 10-yr (t-2) 0.092 (p-value 0.024); Observations 468; R-squared 0.5336.
- LGD treatment
  - Bank-by-bank provisioning rates used as proxy LGDs under the baseline; minimum 50 percent LGD floor on provisioning rate imposed.
  - For IRB portfolios LGDs were stressed by 20 percent in the adverse scenarios.

### Key policy recommendations and priorities
- Solvency
  - Address the sovereign-bank nexus to mitigate interest and credit risk from large bank exposures to the sovereign. (NT)
  - Address household vulnerabilities by imposing a debt service to income ratio (DSTI) limit to mitigate risks of excessive credit growth and rise in defaults. (NT)
- Liquidity
  - Enforce a currency-differentiated liquidity coverage ratio (LCR) for significant currencies. (NT)
  - Monitor a currency-differentiated NSFR for significant currencies. (NT)
- Nonbank Financial Lenders
  - Increase monitoring and ensure provisioning requirements tighten in line with the application of IFRS9 for the banks, to prevent regulatory arbitrage. (NT)
- Time frame notation: C = continuous; I (immediate) = within one year; NT (near term) = 1–3 years; MT (medium term) = 3–5 years.

*Prepared by Maral Shamloo and Armand Fouejieu, Monetary and Capital Markets Department, IMF. Source: cr18163.*

### EXECUTIVE SUMMARY __________________________________________________________________________ 6

### EXECUTIVE SUMMARY

### Introduction and system structure
- The Financial Sector Assessment Program (FSAP) assessed Romania’s financial system, which is largely dominated by banks; 36 banks account for around 80 percent of financial sector total assets.
- The largest five banks concentrate 60 percent of total deposits and 57 percent of total loans.
- Nonbank financial lenders (NBFLs) assets represent about 11 percent of total banking sector assets as of end 2016.
- Bank assets were at just 52.7 percent of GDP as of September 2017.
- Private sector deposits increased from about 48 percent of banks’ total liabilities in 2011 to about 64 percent in 2017.
- Loan-to-deposit ratio fell from 122 percent in 2008 to 75 percent in 2017.
- The share of foreign exchange (FX) denominated loans: 43 percent for corporate loans and 38 percent for household loans in June 2017.
- NPLs declined to 6.4 percent as of December 2017 from their peak of 21.5 percent following the crisis.
- Average banking sector return on assets has grown to 1.5 percent.

### Emerging vulnerabilities identified
- Rapidly increasing bank exposures to the government: sovereign debt exposure rose from less than 5 percent in 2008 to about 22 percent of banks’ total assets in 2016.
- Growing and riskier real estate exposure: housing loans increased from 21 percent of loans to households to more than 54 percent between 2008 and 2017.
- Large stock of FX loans (declining share but still significant) poses credit risk and potential liquidity needs in foreign currency.
- Rapid growth of credit by NBFLs, combined with relatively thin capital levels and riskier lending practices.
- Long duration of domestic sovereign debt relative to funding duration exposes banks to interest rate risk.
- Prima Casa government-backed mortgage program increases indirect banking sector exposure to the public sector and allows for LTV ratios of up to 95 percent, undermining macroprudential effectiveness.
- Majority of mortgage contracts are at variable rates, raising vulnerability to rising interest rates.

### Stress testing scope and scenarios
- Tests assessed banking-system resilience to solvency and liquidity shocks and contagion risks.
- Interest rate risk and credit risk on sovereign securities assessed under a scenario of widening credit spreads and sharply rising interest rates.
- Household vulnerabilities tested through a scenario with significant contraction of domestic output.
- Liquidity ability tested in all major currencies via a battery of liquidity stress tests.
- NBFL sector stresses were also examined.
- Adverse scenario parameters included downside external risks: tightening in global financial conditions, capital outflows, widening sovereign spreads, faster-than-anticipated monetary policy normalization, and a domestic demand shock.
- The adverse scenario produces a V-shaped recession with a cumulative decline of GDP equal to two standard deviations relative to the baseline and a cumulative peak to trough GDP contraction of 8.1 percent.

### Solvency stress test results — banking sector
- Starting point: CET1 ratio for the 12 banks considered was 17.1 percent at end-2016.
- Under the adverse scenario, CET1 declines to 8.2 percent at end-2020 for the 12 banks in scope.
- Main contributors to capital decline under adverse scenario:
  - Credit risk: negative contribution of 750 bps in capital over the horizon.
  - Trading losses on sovereign securities portfolios: 300 bps drop in capital over the horizon.
  - Exchange rate depreciation: important role in driving credit losses on FX loans.
  - Net interest income (NII) decreases by almost 40 percent throughout the test horizon.
- Under the baseline scenario, CET1 for the 12 banks declines from 17.1 percent at end-2016 to 14.1 percent at end-2020.
  - Drivers under baseline:
    - Banks extend credit in line with nominal GDP growth (projected strong growth).
    - NII is on average lower by 20 percent throughout the horizon compared to 2016.
    - Increase in risk-weighted assets (RWAs) contributes a 150 bps decrease in capitalization ratios.
    - Trading losses contribute a further 250 bps drop.
- Sensitivity analyses:
  - Simultaneous default of five largest nonfinancial corporate exposures: most (though not all) banks can withstand this concentration shock.
  - A 50 percent drop in the value of government securities causes a majority of banks to fall under minimum regulatory capital requirements, underscoring sovereign exposure risks.

### Liquidity stress test results
- Aggregate liquidity coverage ratio (LCR) is above 200 percent in both euro and leu at baseline.
- LCR remains significantly above 100 percent under stress scenarios simulating dry-ups in retail and wholesale funding respectively.
- Cashflow-based liquidity stress tests show that, with the exception of one institution, all banks could confront persistent and sizable withdrawals without resorting to central bank support.
- All Romanian banks meet Net-Stable Funding Ratio (NSFR) ratios above 1 (not a requirement), yet several banks fail to meet LCR and NSFR ratios on a currency-by-currency basis.
- Analysis shows liquidity ratios are flattered by a large stock of government securities; reliance on one type of liquid asset is less helpful in a systemic liquidity event.
- Recommendation emerging: need for currency-differentiated liquidity requirements.

### Nonbank Financial Lenders (NBFLs) findings
- NBFLs are smaller but growing; vulnerabilities concentrated among leasing companies.
- Sensitivity tests for NBFLs assessed resilience to depreciation, increase in borrowing costs, and deterioration in asset quality.
- Leasing companies are particularly vulnerable to currency depreciation due to highly euroized lending portfolios and low starting capital levels.
- Recommendations: increase monitoring of NBFLs and ensure provisioning requirements tighten in line with application of IFRS9 for banks to prevent regulatory arbitrage.

### Interconnectedness and contagion risks
- Domestic interbank exposures are small, limiting scope for transmission of stresses among banks.
- Main channel of cross-border exposure is reliance on parent funding by Romanian banks.
- Reputational risks: failure of affiliated NBFLs could lead to sizable reputational losses and retail deposit runs on parent banks if depositors do not differentiate between entities.
- Simulation suggests that reputational contagion from NBFL failure could be sizeable, supporting the need for closer monitoring and harmonized regulatory requirements with banks.

### Key policy recommendations (Table 1 highlights)
- Solvency
  - Address the sovereign-bank nexus to mitigate interest and credit risk from large bank exposures to the sovereign. (NT)
  - Address household vulnerabilities by imposing a debt service to income ratio (DSTI) limit to mitigate risks of excessive credit growth and rise in defaults. (NT)
- Liquidity
  - Enforce a currency-differentiated liquidity coverage ratio (LCR) for significant currencies. (NT)
  - Monitor a currency-differentiated NSFR for significant currencies. (NT)
- Nonbank Financial Lenders
  - Increase monitoring and ensure provisioning requirements tighten in line with the application of IFRS9 for the banks, to prevent regulatory arbitrage. (NT)

Note: Time Frame notation in table: C = continuous; I (immediate) = within one year; NT (near term) = 1–3 years; MT (medium term) = 3–5 years.

*Prepared by Maral Shamloo and Armand Fouejieu, Monetary and Capital Markets Department, IMF.*

### 6.      FX exposures, while decreasing, continue to be a source of credit and liquidity risk.

### 6.      FX exposures, while decreasing, continue to be a source of credit and liquidity risk.

### FX exposures and loan composition
- Foreign currency denominated loans reached 63 percent of total loans in 2011.
- As of December 2016, the share in the stock was:
  - 44 percent for corporate loans in foreign currency.
  - 42 percent for household loans in foreign currency.
- The share of FX in total new bank lending to the nonfinancial private sector was just below 20 percent in June 2017, with FX loans mainly extended to the corporate sector.
- NBFL lending to the corporate sector is predominantly in FX: 84 percent.
- Going forward, FX lending may pick up again if domestic rates were to rise relative to euro area rates.
- Currency mismatch on banks’ balance sheets is limited, but indirect exposure via unhedged borrowers remains significant.
- Meeting liquidity requirements in FX could become a concern if the share of FX denominated loans remains elevated.
- Loans in foreign currency continue to have a higher non-performing ratio than loans in lei (aggregate comparison referenced but specific ratio figures absent in the excerpt).

### NBFLs, household indebtedness, and contagion risks
- Loans from NBFLs to households increased by 63 percent between June 2016 and June 2017 (compared to 20 percent for banks).
- Concerns raised:
  - Contribution to household indebtedness.
  - Regulatory arbitrage: banks may book loans through subsidiary NBFLs subject to lighter-touch supervision.
  - Contagion of risks from NBFLs to the banking system via direct exposures and reputational channels.

### Objective and scope of the FSAP risk analysis
- Objective: assess the capacity of the banking system to withstand severe but plausible macroeconomic shocks; identify weaknesses and policy priorities to reduce exposures and build capital and liquidity buffers.
- Tests conducted:
  - Solvency exercises: Top-Down (TD) scenario-based stress test and single factor sensitivity analyses (FSAP team).
  - Liquidity stress tests: LCR, NSFR, and cashflow based analysis at aggregate level and by major currency.
  - Contagion analysis: cross-border exposures, domestic interbank exposures, bank-NBFL cross-exposures.
- NBFLs: single factor sensitivity tests on capital to credit losses, exchange rate risk, and increase in funding costs.
- The NBR conducted independent TD stress tests under the IMF macro scenario and arrived at results very comparable to the FSAP team.

### Macroeconomic scenarios (three-year horizon)
- Baseline: strong output growth in 2018 and gradual reversion to potential consistent with WEO and NBR projections.
- Adverse scenario: extreme but plausible—widening sovereign credit spreads, severe currency depreciation in year 1, large hike in domestic policy rates, slowdown in trading partners, and V-shaped GDP contraction peaking at trough in year 2.
- Severity of the adverse scenario: 2.1 standard deviations (defined as deviation of GDP growth from the baseline in terms of historical growth volatility).
- Adverse scenario triggers: risk-off capital outflows, faster than anticipated policy rate normalization, larger fiscal shock or expansionary fiscal policies causing loss of confidence.

- Exogenous shocks calibrated in the NBR’s Quarterly Projections Model (QPM):
  - (i) risk premium shock for private and public sector;
  - (ii) external demand shock affecting Romania's main trading partners;
  - (iii) additional internal demand shock.
- LGDs are stressed under the adverse scenario, implying a drop in the value of collateral.

- Table 2. Romania: FSAP Stress Test Adverse Scenario — Deviation from the baseline (In percentage points; unless specified otherwise)
  - 2018 / 2019 / 2020
  - Real GDP: -6.5 / -14.6 / -17.8
  - Policy interest rate: 2.5 / 1.0 / 0.0
  - Short-term money market spread: 0.5 / 0.0 / 0.0
  - Long-term government bond yield: 4.2 / 4.0 / 3.6
  - Real Effective exchange rate appreciation(-)/depreciation(+): 14.7 / 7.3 / 7.5
  - Nominal exchange rate appreciation(-)/depreciation(+): 7.5 / 1.6 / -0.8
  - Inflation rate (CPI): 1.3 / -1.4 / -4.1
  - Unemployment rate: 1.2 / 1.6 / 1.7
  - Memo:
    - Baseline Real GDP growth (in percent): 4.3 / 3.4 / 2.7
    - Real GDP growth (in percent) under adverse: -2.4 / -5.2 / -0.6
    - Cumulative real GDP growth (from 2017): -2.4 / -7.5 / -8.1
    - Spread of short-term money market rate: -2.0 / -1.0 / 0.0
    - "Severity": 1.5 / 2.1 / 1.9

- Table 3. Romania: FSAP Stress Test Baseline and Adverse Scenarios (In percent; unless specified otherwise)
  - Baseline (2018 / 2019 / 2020)
    - Real GDP Growth: 4.3 / 3.4 / 2.7
    - Policy interest rate: 2.3 / 3.3 / 4.4
    - Short-term money market spread: 0.0 / 0.0 / 0.0
    - Long-term government bond yield: 3.7 / 3.9 / 3.8
    - Nominal exchange rate app(-)/dep(+): 0.9 / -0.1 / -0.2
    - Inflation rate (CPI): 3.5 / 3.2 / 2.8
    - Unemployment rate: 5.1 / 5.1 / 5.2
  - Adverse (2018 / 2019 / 2020)
    - Real GDP Growth: -2.5 / -5.6 / -1.2
    - Policy interest rate: 4.8 / 4.3 / 4.4
    - Short-term money market spread: 0.5 / 0.0 / 0.0
    - Long-term government bond yield: 7.9 / 7.9 / 7.4
    - Nominal exchange rate app(-)/dep(+): 8.4 / 1.5 / -1.0
    - Inflation rate (CPI): 4.8 / 1.8 / -1.3
    - Unemployment rate: 6.3 / 6.7 / 6.9

### Top-Down solvency stress test methodology and assumptions
- Coverage: Romania’s 12 largest banks, accounting for 80 percent of banking sector assets (includes nine OSIIs).
- Risks covered: credit risk (all exposures), market risk (mark to market securities, exchange rate, other mark to market positions), interest rate risk in the banking book.
- Balance sheet and RWA growth projections:
  - Quasi-static balance sheet approach: asset allocation and funding composition remain the same; balance sheets grow in line with the nominal GDP path in the scenario.
  - Rate of change of the balance sheet floored at zero percent to prevent deleveraging; this constraint binds in the adverse scenario.
  - Under the baseline, growth of assets calibrated to half of nominal GDP growth.
- RWAs:
  - Standardized portfolios: RWA change due to portfolio size change and increased risk weight for newly defaulted exposures; average risk weight for the remainder maintained.
  - IRB portfolios: projected through-the-cycle (TTC) PDs (calculated as one-eighth of point-in-time (PIT) PDs) used to calculate new average risk weights.
- Modeling and behavioral assumptions:
  - Interest income and expense projected via maturity gap analysis; banks assumed not to change maturity profile over the stress period.
  - Evolution of cost of funding depends on policy rate plus risk-premium in adverse scenario; lending rates tied to the Lombard rate by regulation, preventing full pass-through of funding cost increases.
  - Trading income and losses on securities assessed using portfolio-level information and the modified duration approach for fixed income securities in HFT and AFS portfolios; rebalancing of the portfolio not allowed.
  - HTM securities: provisions made according to changes in credit spreads.
  - Other market positions (commodity and currency net open position) impacted via scenario variables; other income statement items projected to grow with the balance sheet.
- Capital action assumption:
  - Banks do not issue new shares or make repurchases during the stress test horizon.
  - Dividends paid out of current period net income after taxes by profit-making banks at a rate of 10 percent and in compliance with supervisory capital requirements.
- Hurdle rates (Table 4: In percent)
  - Baseline (OSII / N-OSII):
    - Total Capital: 11.5 / 10.5
    - Tier 1 Capital: 9.5 / 8.5
    - CET1: 8.0 / 7.0
    - Basel III regulatory minimum: 4.5 / 4.5
    - Capital conservation buffer: 2.5 / 2.5
    - OSII buffer: 1.0 / 0.0
  - Adverse (OSII / N-OSII):
    - Total Capital: 9.0 / 8.0
    - Tier 1 Capital: 7.0 / 6.0
    - CET1: 5.5 / 4.5
    - Basel III regulatory minimum: 4.5 / 4.5
    - Capital conservation buffer: 0.0 / 0.0
    - OSII buffer: 1.0 / 0.0
  - Assessment: final capital level calculated on a Fully Loaded basis; total capital adequacy ratio evaluated at CET1 requirement plus 3.5 percent; Tier 1 ratio at CET1 requirement plus 1.5 percent; leverage ratio assessed against the 3 percent Basel III minimum. Banks ending the horizon below relevant hurdle rates or leverage ratio are considered to have failed the test.

### Credit risk estimation and loan/securities composition
- Loans are more than half of total banking sector assets: 57 percent.
- Debt securities constitute 22 percent of assets, more than three quarters of which are marked to market.
- Loan composition:
  - Mortgages: 51 percent of the loan book.
  - SME loans backed by real estate: 23 percent of the loan book.
- Debt securities composition:
  - Domestic sovereign securities form 97 percent of total securities holdings of the 12 banks.
- Regulatory approaches:
  - Ten out of 12 banks apply exclusively the standardized approach to credit risk.
  - Two banks apply the IRB approach partially (86 and 67 percent of their respective credit RWAs).
  - Overall, more than 80 percent of credit RWA are under the standardized approach, implying a relatively high average risk weight compared to European peers.
- Default rates estimation:
  - Historical default rates provided by the NBR for four portfolios: retail mortgages, SMEs backed by RE, non-RE backed SMEs, and consumer loans.
  - Point-in-time PDs projected using regression models with macro variables as independent variables; FSAP team estimated PDs using four panel regressions (one for each portfolio).

*Source: cr18163 - 6.      FX exposures, while decreasing, continue to be a source of credit and liquidity risk.*

### 29.      Default rates were estimated using a Bayesian Model Averaging (BMA) approach.

### 29.      Default rates were estimated using a Bayesian Model Averaging (BMA) approach.

### PD estimation methodology
- Default rates were estimated using a Bayesian Model Averaging (BMA) approach.
- Under this methodology:
  - A subset of all possible models is chosen where all explanatory variables (macrofinancial variables and lags) are statistically significant in explaining changes in PDs.
  - The coefficients were then obtained using a weighted average of default rate estimates across multiple models, with the weights corresponding to the posterior probability of each specification (see Appendix III).

### Treatment of large nonfinancial corporate exposures
- PDs for large nonfinancial corporate exposures were set at historically stressed levels.
- Rationale:
  - Large nonfinancial corporate exposures are “low default portfolios” and thus macroeconomic variables have little explanatory power for the variations in these PDs.
  - To overcome this problem, the 80th percentile default rate experienced in the historical data was used for each bank.

### Empirical balance-sheet and loan characteristics (selected figures reported)
- New loans variable-rate shares (loan flows January–September 2017):
  - 88 percent of new loans in lei have a variable interest rate.
  - 85 percent of new loans in euro have a variable interest rate.

### Impact of exchange rate depreciation and government bond yields on credit losses
- Corporate PDs are sensitive to real GDP and long-term government bond yields.
- Retail portfolios are sensitive to real GDP and exchange rate movements.
- Two hypothetical adverse scenarios (identical to the adverse scenario except for exchange rate and long-term government bond yields) were constructed to quantify impacts:
  - First hypothetical: government bond yields remain at their end-2017 levels.
  - Second hypothetical: exchange rate remains at its end-2017 level.
- Key quantified impacts:
  - Credit losses in SME portfolios are 45 percent larger in the presence of a rise in government bond yields (actual adverse scenario compared to the first hypothetical scenario).
  - Mortgage losses are 20 percent higher due to depreciation (actual adverse scenario compared to the second hypothetical scenario).
  - Consumer losses are 16 percent higher due to depreciation (actual adverse scenario compared to the second hypothetical scenario).

### LGD modeling and assumptions
- For modeling LGDs bank-by-bank provisioning rates were used.
- In the absence of time series data on LGDs:
  - The provisioning rate on newly defaulted loans was used as the LGD for each bank under the baseline as a proxy for LGDs.
  - A minimum 50 percent LGD floor on provisioning rate was imposed.
  - For IRB portfolios the LGDs were stressed by 20 percent in the adverse scenarios.

### Market risk (securities portfolios and interest-rate risk)
- The solvency stress test assessed valuation losses in securities portfolios:
  - The value of the debt securities portfolio among the 12 major banks amounts to around RON 71 billion or 22 percent of total assets; 97 percent of which are domestic bonds.
  - Average duration of debt securities portfolio is 3.5 years, with significant variation across banks; banks with longer average duration experienced higher valuation losses.
- Market risk from equity, commodities, and direct exchange rate positions is negligible:
  - Romanian banks do not carry material open positions in equities or commodities; market risk is mostly due to fair-valuation effects on the securities portfolio.
- Repricing gap analysis (interest rate risk on the banking book):
  - Romanian banks have significant positive interest rate risk on their balance sheet, exposing them to losses in interest income as the lending yield curve flattens under the baseline and adverse scenarios.
  - Under the adverse scenario long-term government yields and borrowing costs for the banks rise significantly, while lending rates are constrained as banks’ lending rates are tied to the Lombard rate.

### Solvency stress test results — aggregate outcomes and drivers
- Aggregate outcomes for the 12 banks:
  - Aggregate CET1 ratio declines from 17.1 percent at end-2016 to 8.2 percent at end 2020 in the adverse scenario.
  - Leverage ratio (T1 Capital to Total Assets) decreases from 9.3 to 4.8 percent during the stress test horizon in the adverse scenario.
  - Under the baseline scenario the CET1 ratio declines from 17.1 percent at end-2016 to 14.1 percent at end-2020; leverage ratio decreases from 9.3 to 7.7 percent.
- Main contributors to capital depletion in the adverse scenario:
  - Trading losses: total drop of 300 bps in capital over the three-year horizon; trading gains of 0.6 percent of RWA in 2017 turn to a market loss of 3.6 percent in 2018.
  - Credit risk: total drop of 750 bps in CET1 during the three-year horizon; credit loss impairments increase from -2.2 percent of RWA in 2016 to -3.5 percent of RWA in 2019 (peak provisioning).
    - Composition of credit losses: real estate mortgages account for 54 percent of total; SME backed by real estate account for 20 percent of total.
  - Net interest income (NII): decreases by almost 40 percent through the stress test horizon.
    - Contribution of NII to CET1 is, on average, 280 bps per year throughout the horizon, compared to 525 bps in 2017.
    - Drivers: large NPL formation compresses interest income; NIM adversely affected due to rise in funding costs not matched by lending rates tied to the Lombard rate; positive interest risk exposes banks to losses as ROBOR increases.
- Under the baseline scenario additional contributors:
  - Trading losses contribute a 250 bps drop in capital levels.
  - An increase in RWA contributes a 150 bps decrease in capitalization ratios due to balance sheet growth assumptions (banks’ balance sheets assumed to grow at half of nominal GDP growth).
  - Banks’ NII is on average lower by 20 percent compared to 2016 due to strong increase in short term rates and positive interest rate risk.

### Stress test tables and summary metrics (selected)
- Table 5. Banking Sector Solvency Stress Test Results (In percent) — aggregate outcomes:
  - Before stress: Aggregate CET1 Ratio 17.1; Leverage Ratio 9.3; Asset Share Undercapitalized (CET1<Hurdle*) 0.0; Asset Share Undercapitalized (T1<Hurdle*) 0.0; Asset Share Undercapitalized (CAR<Hurdle*) 0.0; Asset Share Undercapitalized (leverage ratio<3%) 0.0; Maximum Capital Shortfall (% of GDP) 0.0.
  - Baseline scenario: Aggregate CET1 Ratio 14.1; Leverage Ratio 7.7; Asset Share Undercapitalized (CET1<Hurdle*) 0.0; Asset Share Undercapitalized (T1<Hurdle*) 13.1; Asset Share Undercapitalized (CAR<Hurdle*) 13.1; Asset Share Undercapitalized (leverage ratio<3%) 0.0; Maximum Capital Shortfall (% of GDP) 0.1.
  - Adverse scenario: Aggregate CET1 Ratio 8.2; Leverage Ratio 4.8; Asset Share Undercapitalized (CET1<Hurdle*) 30.3; Asset Share Undercapitalized (T1<Hurdle*) 30.3; Asset Share Undercapitalized (CAR<Hurdle*) 28.3; Asset Share Undercapitalized (leverage ratio<3%) 30.3; Maximum Capital Shortfall (% of GDP) 0.3.

### Solvency sensitivity analysis (single-factor tests)
- A set of sensitivity tests assessed vulnerabilities to sovereign, interest rate, and concentration risks:
  - A 500 bp increase in sovereign yields reduces the system Tier I capital by 510 bps; impact can be up to 970 bps for certain banks.
  - A 500 bp increase in interest rates reduces the system Tier 1 ratio by 120 bps, with larger impacts for some banks.
  - Concentration risk: most, but not all, Romanian banks can withstand the simultaneous default of five of their largest Nonfinancial Corporate (NFC) exposures; collateral is mainly cash or T-bills; CRE collateral is not accepted as a credit risk mitigant for large exposures in Romania.
  - A 50 percent haircut on sovereign exposures would cause several banks to fall below the 8.5 percent regulatory minimum on Tier 1 capital, demonstrating very large sovereign exposures for a majority of banks.
- Additional sensitivity tests to exchange rate and commodity shocks were performed but had negligible impact because Romanian banks do not carry material open positions in equities or commodities.

### Liquidity stress tests (overview)
- Three liquidity stress tests were conducted for the 12 banks:
  - Liquidity Coverage Ratio (LCR) and Net Stable Funding Ratio (NSFR) analyses, conducted separately in lei and euro.
  - Cashflow-based analysis simulating outflows over maturity buckets from 1 day to 365 days (supervisory liquidity reporting templates used).
- The cashflow-based analysis used supervisory data from local supervisory liquidity reporting templates provided by the NBR.

*Source: IMF staff calculations, from the FSAP chapter text.*

### 41.      The structure of banks’ funding appears relatively homogenous across the main

### 41.      The structure of banks’ funding appears relatively homogenous across the main

### Funding structure and Available Stable Funding (ASF)
- At the aggregate level for the 12 largest Romanian banks in scope, the funding structure in the two main currencies (lei and euro) is very similar.
- Funding composition (aggregate):
  - Retail funding: 52 percent
  - Non-operational deposits: 23 percent
  - Other unsecured funding: 18 percent
  - Secured funding: almost non-existent
- ASF composition:
  - 56 percent of lei ASF is met by retail deposits
  - 60 percent of euro ASF is met by retail deposits
  - Funding from financial institutions is the next largest ASF component after retail deposits

### High Quality Liquid Assets (HQLA) composition
- HQLA are dominated by government securities in both lei and euro.
- HQLA composition (by exposure):
  - Central government exposures:
    - 78 percent in local currency (RON)
    - 65 percent in euro (EUR)
  - Central bank assets (including excess reserves):
    - 8 percent in local currency (RON)
    - 19 percent in euro (EUR)
  - Coins and banknotes and Multilateral and int. Org. assets are present in smaller shares as per Figure 10

### Maturity profile and interest rate risk
- Available stable funding maturity profile:
  - 65 percent of available funding has a maturity of less than six months
  - Only a quarter of funding is of maturity greater than 12 months
- Required stable funding (RSF) maturity profile:
  - 80 percent of required funding is of maturity of greater than 12 months
- Implication:
  - Large mismatch in maturity structure of funding contributes to large positive interest rate risk observed in the solvency stress test

### B. LCR-Based Liquidity Stress Test — Scenarios implemented
- Basel III prescribed scenario applied.
- Two additional FSAP scenarios to assess abrupt withdrawal of funding:
  - Retail stress scenario (deposit run replication):
    - Run-off rates: 10 percent for stable retail deposits, 15 percent for unstable retail deposits
  - Wholesale funding stress scenario:
    - Complete loss of wholesale funding on the interbank market
    - Run-off rates of 100 percent for wholesale funding from other legal entity customers
    - Run-off rates of 50 percent for operational deposits generated by clearing, custody, and cash management activities
    - Outflows by nonfinancial corporates, central banks and multilateral development banks up to 50 percent
- In both additional scenarios:
  - A 15 percent haircut is applied to government debt to simulate a fire sale or illiquid market for these securities
  - A detailed table of all haircuts is included in Appendix IV (referenced in source)

### B. LCR-Based Liquidity Stress Test — Results and sensitivities
- Aggregate LCR:
  - Aggregate LCR is above 200 percent in both euro and lei, significantly above the 100 percent Basel III benchmark
- Currency-by-currency issues:
  - Some banks meet aggregate LCR while missing the 100 percent benchmark on a currency-by-currency basis
  - Specifically, a number of banks fall short of the RON 100 percent requirement and some banks fall short in euro
- Sensitivity to government securities haircut:
  - Results are highly sensitive to haircut assumptions on government securities because they form the largest portion of HQLA
- Under stressed scenarios, several banks do not meet 100 percent LCR on a currency-by-currency basis
- Average liquidity shortfall:
  - Ranges from 1 to 3 percent of banks' total assets (average)
- Table 6 summarized system-wide and scenario LCRs and shortfalls (preserving reported numbers):
  - System-wide LCRs:
    - RON (Basel III): 233%
    - RON (Retail Scenario): 184%
    - RON (Wholesale Scenario): 176%
    - EUR (Basel III): 203%
    - EUR (Retail Scenario): 167%
    - EUR (Wholesale Scenario): 159%
  - Liquidity shortfall (RON mil):
    - RON Basel III: 785
    - RON Retail Scenario: 1,925
    - RON Wholesale Scenario: 2,384
    - EUR Basel III: 362
    - EUR Retail Scenario: 613
    - EUR Wholesale Scenario: 816
  - Share of assets (average) corresponding to the shortfalls:
    - RON Basel III: 1%
    - RON Retail Scenario: 2%
    - RON Wholesale Scenario: 3%
    - EUR Basel III: 2%
    - EUR Retail Scenario: 2%
    - EUR Wholesale Scenario: 2%

### C. NSFR-Based Liquidity Stress Test — Findings
- Methodology:
  - NSFR ratios calculated for each bank on a currency-by-currency basis using Basel III weights and banks’ EBA NSFR template reporting
  - NBR does not yet calculate or monitor an actual NSFR requirement domestically
- Results:
  - Almost all banks meet an NSFR ratio of above one in euro and lei
  - Only one (small) bank fails to meet the NSFR on a euro basis
  - NSFR ratios are flattered by the large presence of government securities, which have a low weight in terms of required funding
  - This is evidenced by comparing average RSF factor for Romanian banks to a representative sample of European banks analyzed by the EBA

### D. Cashflow-Based Liquidity Stress Test — Maturity ladder (TD) approach and results
- Methodology:
  - Cash-flow based liquidity stress tests implemented through a TD approach using supervisory information on contractual maturity structure of assets and liabilities
  - Tests assessed outflows, inflows, and available buffers; applied run-off rates by funding type and asset valuation haircuts
- Results:
  - Except for one small bank, all banks have sufficient counterbalancing capacity (cash, excess central bank reserves, or government bonds) to meet outflows
  - In line with haircuts applied to eligible assets for NBR standing facility provision, all government securities are accepted at zero haircut (assumption that banks can convert government securities to finance outflows through NBR’s standing facilities)

### Stress test of Nonbank Financial Lenders (NBFLs) — Introduction and sector structure
- Rationale for NBFL assessment:
  - NBFL assets are almost 11 percent of the banking sector in total and rapidly expanding
  - NBFLs do not undergo regular stress testing by the NBR
  - Significant linkages with the banking sector (many NBFLs consolidated by local institutions and receive funding from parents) pose contagion risk
  - NBFLs likely have less sophisticated risk management and corporate governance
- Scope and composition:
  - Analysis limited to 35 firms in the special register
  - 22 of the 35 NBFLs in scope are leasing firms, representing about two-thirds of total assets in the sector
  - Remaining firms specialize mainly in consumer (uncollateralized) lending
- Balance sheet composition differences:
  - Leasing companies:
    - Liabilities: Loans from banks represent 76 percent of total liabilities (figure captioning shows Loans from Banks (76%))
    - Assets: Fixed assets around two-thirds (66%); credit to customers 22%; claims against financial institutions 5%; others 7%
    - Credit to customers: 96 percent to SMEs; close to 90 percent in foreign currency; only 4 percent to individuals
    - Average interest rate on loans: around 5 percent; average interest margin: 2 percent
    - Capital: 11 percent (closer to minimum requirement)
    - NPLs: 13 percent; coverage ratios reported as comfortable but include general provisions for performing loans
  - Non-leasing companies:
    - Liabilities: Loans from banks represent 87 percent of total liabilities (figure captioning shows Loans from Banks (87%))
    - Assets: Credit to customers 63%; claims against financial institutions 18%; fixed assets 1%; others 18%
    - Credit to customers: 67 percent to individuals (95 percent denominated in domestic currency); credit to SMEs 33 percent split roughly equally in domestic and foreign currency
    - Average interest rate on loans: 12 percent; average interest margin: 9 percent
    - Capital: 37 percent (comfortably above minimum requirement)
    - NPLs: 22 percent

### NBFL stress testing — Sensitivity tests and assumptions
- Single-factor sensitivity tests for the 35 firms assessed:
  - A 10 percent depreciation of the domestic currency:
    - Captures direct impact through net open position and indirect impact via FX-denominated loan losses
    - Loss parameters for NBFLs calibrated using coefficient on exchange rate movement in PD estimation for SME loans from banks (due to lack of NBFL time-series defaults)
  - A five percent increase in NPLs:
    - Additional provisions associated with a 5 percent increase in NPLs deducted from capital of each firm
    - A 100 percent coverage assumed (in line with observed coverage for NPLs)
  - A 300 basis points increase in the cost of funding:
    - Additional cost of borrowing deducted from existing capital

### NBFL stress testing — Results and vulnerabilities
- Overall finding:
  - Leasing companies are particularly vulnerable due to highly euroized lending portfolio and low capital levels
- Specific outcomes:
  - A 10 percent depreciation of the domestic currency:
    - Aggregate capital level for leasing companies decreases from 11.5 percent to -1.2 percent
    - Several firms accounting for 82 percent of the subsector’s assets fall under the 7 percent regulatory minimum capital (note: number of failures reported is in addition to two firms already undercapitalized)
    - Almost all impact comes from indirect loan losses due to depreciation
    - Non-leasing companies: capital decreases to 34 percent from 37 percent; only 1 firm falls below regulatory minimum
  - A five percent increase in NPLs:
    - For leasing companies: capital ratio falls by 430 bps to an aggregate of 7.2 percent; several firms accounting for 72 percent of total assets fall below minimum capital requirement
    - For non-leasing companies: impact is 330 bps, bringing sectoral capital down to 33 percent; no firm falls below regulatory minimum
  - A 300 bps increase in cost of funding:
    - For leasing firms: capital to assets ratio reduced by 200 bps; several institutions representing around 50 percent of the sub-sector by asset share fall below minimum capital requirement
    - For non-leasing firms: impact of 170 bps; no firm falls below minimum due to higher initial capital levels
- Conclusion:
  - Vulnerabilities are related to low capital levels of the leasing subsector and their highly euroized balance sheets
  - A more tailored estimate of loan losses for NBFLs, similar to those conducted for banks, is warranted given sensitivity to exchange rate depreciation if borrowers are not largely naturally hedged

### Interconnectedness analysis and contagion risks — Overview
- Objective:
  - Explore risk of contagion of stresses through balance-sheet exposures domestically and abroad
- Domestic analysis:
  - Assessed interbank and bank-NBFL exposures using institution-level data to examine contagion impact of failures on the network
- Cross-border analysis:
  - Explored Romanian banking system’s exposures to shocks from foreign banking sectors
- Relationship to Balance Sheet Approach:
  - Contagion analysis complements the Balance Sheet Approach; the latter assesses exposures among subsectors (banks, investment funds, insurance companies, pension funds, etc.) and how cross exposures have changed over time
  - Contagion analysis focuses on banks and NBFLs and uses institution-level data

*IMF staff calculations and FSAP analysis as presented in the source document.*

### 61.      The domestic interconnectedness analysis investigated channels of contagion between

### The domestic interconnectedness analysis investigated channels of contagion between

### Domestic interbank contagion: scope and methodology
- Analysis focused on contagion between one or a group of institutions and the rest of the financial system given the relative concentration of the banking system in Romania.
- Network: top 12 banks by share of banking sector assets (remaining banks treated as a single entity). Data as of end-2016.
- Methodology: network approach developed by Espinoza-Vega and Sole (2011).
- Contagion calibration for the interbank stress-test scenario:
  - 100 percent loss of claims in case of default.
  - 65 percent rollover of funding.
  - 50 percent haircut on assets subject to liquidation.
- Fire-sale channel modeled: affected banks replace a fraction of lost funding by selling assets at a discount, allowing domino effects until no further failures.

### Domestic interbank contagion: results
- Index of contagion (outward measure: average loss as a share of capital for other banks due to failure of one bank) ranges from 0.01 to 1.6 percent.
- Index of vulnerability (inward measure: losses as a share of capital for a bank due to failure of the rest of the network) ranges from 0.07 to 1.1 percent.
- The bank with the highest outward contagion is also the most vulnerable to distress from the rest of the system.
- Conclusion: low level of losses points to a limited risk of contagion through interbank exposures.

### Bank–NBFL interconnectedness: importance and scope
- NBFLs are gaining importance: stock of loans to the private sector granted by NBFLs reached RON 26.6 billion in September 2017, representing around 10.3 percent of total loans to households and nonfinancial corporations.
- Significant linkages with the banking system: many NBFLs are consolidated by local institutions.
- Network analysis covered the largest banks and NBFLs:
  - Interbank and bank-NBFL gross credit exposures assessed on both asset and liability sides.
  - For bank–NBFL contagion the largest 11 banks and 34 largest NBFLs (data as of end-2016) were considered.
- Two NBFL scenarios:
  - Scenario 1: vulnerabilities of the banking sector to failure of each NBFL (credit shock only; 100 percent loss in case of default).
  - Scenario 2: vulnerabilities of NBFLs to stresses in the banking sector (credit shock plus funding shock; 50 percent haircut on assets subject to liquidation; 65 percent rollover of funding).

### Bank–NBFL contagion: results (individual and aggregate)
- Individual NBFLs can have greater systemic impact than individual banks because banks’ exposures to NBFLs (as a share of capital) are sometimes larger than interbank exposures.
- Index of contagion (outward systemic impact) for banks to NBFL failures: broadly in line with interbank results, ranging from 0 to 0.2 for all but one bank.
- Index of contagion for NBFLs (outward systemic impact) ranges between 0.2 and 1.2.
- In Scenario 2 (credit + funding shock) the index of contagion for banks is higher relative to Scenario 1, indicating banks’ failure has larger consequences for NBFLs when funding shocks are included.

### Bank–NBFL contagion: sector‑level (two-entity) results
- Treating 11 banks and 34 NBFLs as two single entities:
  - Scenario 1 (credit shock): index of contagion for NBFLs = 8.6 (banks incur losses of 8.6 percent of their aggregate capital in case of failure of all NBFLs).
  - Scenario 1 (credit shock): index of contagion for banks = 1.7 (NBFLs’ aggregate losses in case of failure of the banking system will amount to 1.7 percent of the NBFL sector’s capital).
  - Scenario 2 (credit + funding shock): index of contagion = 50 for banks, implying that in case of failure of the banking system, NBFLs will lose about 50 percent of their total capital.
- Interpretation: higher exposure on the assets side for banks, and on the liability side for NBFLs.

### Direct and indirect impact of NBFL failure on parent banks
- Six banks own fully or partially an “own-name” NBFL with capital participations ranging from 50 to 100 percent of the NBFL’s total capital.
- Parent bank vulnerabilities from NBFL failure:
  - Direct exposure losses: entire capital participation assumed wiped out in case of NBFL default.
  - Lending to the affiliated NBFL: assume 100 percent loss given default for any funding to the affiliate.
  - Reputational losses: assumed retail deposit runoff of 20 percent; replacement of lost deposits:
    - 50 percent replaceable with other funding for foreign banks.
    - 30 percent replaceable for domestic banks.
    - Remaining financing needs met by selling assets with a 50 percent haircut.
- Calibration of LGDs aligned with contagion analysis assumptions; run-off rates aligned with the liquidity stress test (retail stress scenario).
- Quantified impact:
  - Capital ratios decline by 3 to 9 percentage points for parent banks.
  - Liquidity ratios drop by more than half for one bank.
- Finding: reputational risks can be far more significant than direct exposures as a vulnerability channel, but high initial capital and liquidity imply banks would remain resilient to such shocks.

### Cross-border contagion: exposures and methodology
- Foreign banks represent the majority of Romania’s banking system (27 out of 37 banks).
- BIS locational data (data as of July 2016) used to assess cross-border exposures:
  - Romania’s banking system claims against banks abroad account for about 4 percent of total Romanian banking sector assets.
  - Liabilities to foreign banks represent about 15 percent of total assets.
  - The Austrian banking sector is the top foreign counterpart both on claims and liabilities.
- Cross-border contagion exercise considered 15 banking sectors (Austria, Australia, Belgium, Canada, Cyprus, Denmark, Finland, France, Greece, Italy, Luxembourg, Romania, Switzerland, United Kingdom, and the United States).
- Two scenarios for cross-border contagion:
  - Scenario (i) funding shock only: 30 percent roll-over of funding; 50 percent haircut on assets subject to liquidation.
  - Scenario (ii) combined credit and funding shocks: 100 percent loss on claims in case of default; same rollover and haircut calibrations as scenario (i).

### Cross-border contagion: results
- Scenario (i) funding shock:
  - Index of contagion (outward systemic impact) ≈ 0.01 percent.
  - Index of vulnerability (inward systemic impact) = 10 percent.
- Scenario (ii) combined credit and funding shocks:
  - Index of contagion = 0.08 percent.
  - Index of vulnerability = 15 percent.
- Conclusion: distress in Romania’s banking sector would have very little impact on other banking sectors on average, but the Romanian banking system is relatively more vulnerable to external shocks, particularly via foreign funding.

*International Monetary Fund, Romania Financial System Stability Assessment (selected section).*

### CONCLUSION

### CONCLUSION

### Integration with Macroprudential Workstream
- The findings in the risk analysis workstream provide analytical underpinnings for recommendations in the macroprudential workstream (paragraph 76).
- Results in this note are linked with the Macroprudential Technical Note and its policy recommendations (paragraph 76).

### Recommendations: Sovereign‑Bank Nexus (paragraph 77)
- Findings:
  - Mark-to-market losses associated with securities portfolios account for a large share of the drop in capitalization ratios under the adverse scenario as credit spreads widen and domestic interest rates rise against depreciation pressures.
  - Repricing of securities contributes negatively to capitalization ratios even under the baseline scenario, as growth gains momentum and monetary policy normalizes.
- Recommendation:
  - Address risks arising from large exposures to the government due to (i) potential re-emergence of sovereign risks and (ii) interest rate risk as monetary policy normalizes.

### Recommendations: Liquidity Requirements and Monitoring (paragraph 78)
- Findings:
  - Aggregate liquidity is ample.
- Recommendations:
  - Enforce LCR requirements on a currency-by-currency basis.
  - Start monitoring the NSFR to provide an accurate picture of maturity mismatches in banks’ balance sheets.

### Recommendations: Non‑Bank Financial Institutions (NBFLs) (paragraph 79)
- Findings:
  - Analysis highlights vulnerabilities in the NBFL sector and the potential for stresses in this sector to permeate to banks.
  - NBFLs in the special register are not deposit taking yet can pose contagion risk.
- Recommendations:
  - Closer monitoring of NBFLs.
  - Harmonize provisioning requirements, to the extent possible, between banks and NBFLs for similar loans, in line with the application of IFRS9, to prevent regulatory arbitrage.

### Risk Assessment Matrix — Key Risks and Staff Assessments (Appendix I)
- 1. Re-emergence of financial stress in European banks followed by global risk aversion
  - Likelihood: Staff assessment: Medium
  - Expected impact on financial stability: Staff assessment: High
  - Mechanism: Rise in risk premia on Romanian government bonds exposes banks to losses on domestic government bond positions; adverse effect on economic activity and social tensions.
- 2. Policy uncertainty/divergence and weakening confidence in Romanian economy
  - Likelihood: Staff assessment: High
  - Expected impact on financial stability: Staff assessment: High
  - Mechanism: Higher borrowing costs, potential credit rating impact, sudden capital outflows, slower growth.
- 3. Surges in global financial market volatility associated with exit from unconventional monetary policies
  - Likelihood: Staff assessment: High
  - Expected impact on financial stability: Staff assessment: High
  - Mechanism: Rapid rise in interest rates, steepening yield curve, currency depreciation, higher NPLs, losses on domestic government bond positions.
- 4. Protracted period of slower growth in advanced and emerging economies
  - Likelihood: Staff assessment: Medium
  - Expected impact on financial stability: Staff assessment: Medium
  - Mechanism: Weaker trade/remittances, higher NPLs, lower bank profitability.

### Stress Testing Framework and Key Assumptions (Appendix II)
- Banking sector: solvency risk
  - Exercise: Top-Down by FSAP team.
  - Institutions included: The top 12 banks by share of assets.
  - Market share: 80 percent of total assets in the banking system.
  - Latest data: December 2016.
  - Scope: Consolidated banking group; sovereign and non-sovereign securities exposures covered (held to maturity, available for sale, and fair value accounts), valued at MTM or fair-value at starting point.
  - Stress test horizon: 2018–20.
  - Supporting data note: While data as of December 2016, Q3 2017 unaudited data shows CAR for the banking sector stood at 19 percent, and T1 capital ratio at 17.1 percent.
- Channels and methodology:
  - Macroeconomic scenarios quantified using the NBR’s Quarterly Projections Model (QPM).
  - FSAP team satellite models and balance-sheet regulatory approach.
  - Satellite models estimate credit losses for four portfolios: Consumer, Mortgages, SME backed by RE, SME not backed by RE. PDs for large corporates calibrated to the 80th percentile.
  - Losses from sovereign debt holdings: Haircuts based on a modified duration approach and historical distributions of changes in yield.
- Tail shocks and scenarios:
  - Adverse (V-shaped) scenario: cumulative decline of GDP of two standard deviations over two years, comprising external demand shock, risk premium shock (nominal depreciation), and internal confidence shock.
  - Sensitivity tests include failure of largest 1, 2, 5, and 10 corporate exposures and largest non-corporate exposure; credit, interest, and market risk sensitivities.
- Risks and buffers assessed:
  - Credit losses by exposure type and country.
  - Losses from debt instruments in banking and trading books.
  - Impact of funding cost shocks on net interest income.
  - Market risk, including foreign exchange risk.
- Behavioral adjustments:
  - Static balance sheet.
  - Dividends allowed only for banks that remain adequately capitalized and have positive profits; payout rates consistent with past experience.
- Reporting:
  - Capital shortfall system wide and share in whole system.
  - Dispersion of capital ratios: min., avg., max.; percentage of assets that fail.

### Liquidity Stress Testing (Appendix IV) — Key Elements
- Institutional perimeter:
  - Top 12 banks by share of assets; market share: 80 percent of total assets.
  - Latest data: December 2016; source: supervisory data.
- Methodology:
  - Cashflow-based using time structure of undiscounted cashflows for up to one year.
  - Variants of LCR ratios by currency; NSFR considered.
- Risks assessed:
  - Funding liquidity and market liquidity.
- Scenarios:
  - Retail shock and Wholesale shock.
- Reporting:
  - Liquidity gap by bank.
  - Survival period in days by bank and number of banks that can meet obligations.
- LCR eligibility examples (Scenario types 1/2/3):
  - Coins and banknotes: 100% / 100% / 100%
  - Withdrawable central bank reserves: 100% / 100% / 100%
  - Central government assets: 100% / 85% / 85%
  - Qualifying CIU shares/units: underlying is Level 1 assets excluding extremely high quality covered bonds: 95% / 100% / 100%
  - Extremely high quality covered bonds: 93% / 93% / 100%
  - Qualifying CIU shares/units: underlying is extremely high quality covered bonds: 88% / 88% / 100%
  - Alternative Liquidity Approaches: Inclusion of Level 2A assets recognised as Level 1: 80% / 80% / 100%
- Representative outflow assumptions (over 30 days) — selected entries:
  - Retail deposits where payout agreed within following 30 days: 100% / 100% / 100%
  - Category 1 retail outflows: 15% / 15% / 15%
  - Stable deposits: 5% / 10% / 5%
  - Operational deposits covered by DGS: 5% / 5% / 50%
  - Operational deposits not covered by DGS: 25% / 25% / 100%
  - Non-operational deposits by financial customers: 100% / 100% / 100%
  - Covered by DGS (non-operational other customers): 20% / 20% / 40%
  - Not covered by DGS (non-operational other customers): 40% / 40% / 60%

### Contagion and NBFL Stress Tests (Appendices II–IV)
- Contagion perimeter:
  - Domestic network analysis: top 12 banks by assets; rest of banks treated as single institution.
  - Bank-NBFL contagion: top 12 banks and the 34 NBFLs in the special register.
  - Top 12 banks account for 80 percent of total assets; top 34 NBFLs account for 90 percent of total NBFLs assets.
  - Cross-border analysis: Romania and 14 BIS reporting countries; BIS Locational Statistics: July 2016.
- Methodology:
  - Network analysis using interbank and bank-NBFL exposures (Espinosa-Sole approach).
  - Tail-shock assumptions for domestic analysis: 100 percent loss given default (LGD); 65 percent roll-over of funding; 50 percent haircut on asset fire sales.
  - Tail-shock assumptions for cross-border analysis: 100 percent LGD; 30 percent roll-over of funding; 50 percent haircut on asset fire sales.
- Outputs:
  - Capital shortfall by institution and system-wide, failure of individual institutions, interconnectedness measures.
- NBFLs: solvency sensitivity tests
  - Institutions included: The 34 NBFLs in the special register.
  - Market share: 90 percent of total assets in the NBFLs.
  - Latest data: December 2016.
  - Tests: Sensitivity tests to credit risk and funding risk.
  - Reporting: Losses system-wide and as a share of sectoral assets.
- Policy implication reiterated:
  - Tighten provisioning requirements for NBFLs and harmonize with banks to reduce regulatory arbitrage and systemic spillovers.

### PD Estimation Models (Appendix III) — Approach and Selected Results
- General approach:
  - PDs projected for each bank using historical default rate series for four portfolios: retail mortgages, SME backed by RE, non-RE backed SME, and consumer loans.
  - Data: Quarterly, bank by bank, from Q1 2006 to Q4 2016.
  - Logit transformation applied: Y_it = ln(PD_it / (1 − PD_it)).
  - Logit PDs modeled as linear function of lagged PDs and macrofinancial regressors. Bayesian model averaging (BMA) for panel regressions used to address model uncertainty.
  - Projected logit PDs under stress transformed back to PD space.
- Estimated sensitivities and notes:
  - Corporate PDs mostly sensitive to lagged real GDP growth rate, equity indices and interest rates.
  - Mortgage PDs sensitive to GDP, unemployment and interest rates.
  - Elasticities are small, reflected in relatively small multipliers.
- Selected PD estimation coefficients (Appendix Table 1) — coefficients and p-values:
  - Consumer:
    - PD (t-4): 0.433 (p-value 0.000)
    - D_GDP: -0.075 (p-value 0.000)
    - Constant: -2.454 (p-value 0.000)
    - Observations: 468
    - R-squared: 0.3675
  - Retail Mortgage:
    - PD (t-4): 0.423 (p-value 0.000)
    - D_GDP: -0.108 (p-value 0.000)
    - D_GDP (t-4): 0.020 (p-value 0.004)
    - EXR (t-2): 0.037 (p-value 0.040)
    - 3-month rate (t-4): -0.069 (p-value 0.000)
    - Constant: -2.275 (p-value 0.000)
    - Observations: 468
    - R-squared: 0.5134
  - SME nonRE:
    - PD (t-4): 0.378 (p-value 0.000)
    - D_GDP: -0.062 (p-value 0.003)
    - Constant: -4.174 (p-value 0.000)
    - Observations: 468
    - R-squared: 0.3629
  - SME_RE:
    - PD (t-4): 0.589 (p-value 0.000)
    - D_GDP: -0.076 (p-value 0.000)
    - EXR (t-2): 0.046 (p-value 0.008)
    - Government 10-yr (t-2): 0.092 (p-value 0.024)
    - Constant: -2.114 (p-value 0.000)
    - Observations: 468
    - R-squared: 0.5336

*Source: IMF staff (CONCLUSION and Appendices from the provided document).*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2018/cr18163.pdf_