## _cr11334

## Source details

**Canonical URL:** [_cr11334](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2011/_cr11334.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2011/_cr11334.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2011/_cr11334.pdf.json)

---

### Executive Summary
- Stress testing for this FSAP combined CBR top-down approaches and a bottom-up exercise covering 15 top banks (about 57 percent of the system).
- Top-down tests:
  - Bank-by-bank supervisory data as of end-2010 covering all existing banks in the system (1012).
  - Single-factor tests and macro-scenario tests implemented by the CBR.
- Bottom-up exercise:
  - 15 top banks, covering about 57 percent of the system; shared stress scenarios and macroeconomic assumptions with top-down tests.
- Resilience benchmark: current minimum regulatory capital adequacy ratio (CAR) requirement of 10 percent.
- Risk coverage: credit, concentration, market (foreign exchange, equity, and interest rate), liquidity, and interbank contagion risks; macro scenario test focused on credit, market and liquidity with one year horizon.
- Severe macro scenario: impact equivalent to 1.7 standard deviation shock on real GDP growth.

### Key stress test results — system-level
- Overall resilience: system broadly resilient to a variety of macroeconomic and financial shocks.
- Capital and losses:
  - Gross losses in a tail economic event: about 35 percent of capital (mainly credit losses).
  - About one-third of these losses would be offset by profits.
  - System as a whole would maintain a 14 percent capital ratio, but banks representing about 8 percent of the system (mainly large private banks) would fall below the 10 percent minimum capital ratio.
- Liquidity risk:
  - Acute, systemic liquidity shocks (including FX liquidity) may burden the banking sector significantly in a very short period.
  - Funding would be very volatile during a stress, including individual deposits.
  - Stress tests do not incorporate CBR policy reaction to liquidity shocks.
- Market and FX risk:
  - Valuation losses on securities, especially bonds, could be notable.
  - Direct FX valuation risk negligible given small open FX positions.
- Concentration risk: significant, especially for smaller banks, given high credit concentration.
- Bottom-up vs top-down: broadly similar results.

### Structural and supervisory weaknesses (adjustments and impact)
- Adjustments for vulnerabilities can make the system more fragile than stress tests suggest; baseline balance sheet and capital adjustments dwarf most economic shocks.
- Specific adjustment effects:
  - Adjusting for potential overestimation of loan quality and low provisions could give much larger shocks to capital.
  - Adjusting provisions as described could wipe out as much as one-third of the capital.
  - Adjusting for credit quality of extended maturity loans can reduce capital by over 20 percent, especially for state-owned and large private banks.
- Data caveat: magnitude of adjustments driven in part by subjective assumptions in absence of hard data.

### Mitigating factors and policy capacity
- High capital buffers and strong profitability can act as shock absorbers.
- Fiscal and external capacity:
  - Estimated economic cost of recapitalization is small and manageable owing to the relatively small size of the banking sector relative to GDP, low government debt, and high reserves.
- Policy precedent: government interventions to support and recapitalize weak banks may increase moral hazard but availability of resources mitigates systemic risk.

### Financial sector structure and recent performance
- Size and composition (end-2010):
  - Total assets of financial institutions: around 80 percent of GDP.
  - Banks accounted for over 90 percent of total assets.
  - Bank deposits: just above 40 percent of GDP.
- Number and concentration:
  - 1,012 banks operating in Russia (end-2010), down by 46 from January 1, 2010, and about 250 less than five years earlier.
  - CBR minimum capital requirement raised to Rub 90 million in January 2010, with further increases to Rub 180 million in 2012 and Rub 300 million in 2015 planned.
  - State-owned banks share in total deposits: 52 percent (end-2010).
- Balance sheet composition:
  - Majorities of assets are loans (mostly to industries), then securities (government and corporate bonds, then equities) and interbank transactions (60 percent vis-à-vis non-resident banks).
  - Borrowings from non-residents declined to 13 percents of the book (end-2010), down from 20 percent in 2007, mostly long-term.
  - Liquidity support from the CBR rose substantially in 2008 but was largely withdrawn by end-2010.
- Bank-group performance (post-crisis recovery):
  - Aggregate CAR: 18.1 percent in December 2010.
  - NPL ratio stabilized despite withdrawal of forbearance.
  - Funding improved: household and corporate deposits grew strongly; CBR emergency liquidity support discontinued.
  - Asset growth resumed but slower and skewed toward securities over credit.
  - Profitability rebounded in 2010 though still weak by Russian standards.

### Hidden vulnerabilities and sources of overstated credit quality
- Foreclosed assets often reported at overvalued prices; some collateral quality and saleability in distress not accurately reflected.
- Opaque transfers of distressed assets to off-balance-sheet entities.
- Restructured loans:
  - Share of restructured loans among large loans rose during crisis and remained higher than pre-crisis; more than 90 percent of restructured loans are in performing categories.
  - Loans with extended maturities likely have lower credit quality; no visible change after forbearance withdrawal.

### Stress testing framework, practice, and suggested improvements
- Historical FSAP (2008) used top-down single-factor and bottom-up tests; recommendations included developing macro scenario stress testing, credit risk VaR models, collecting effective maturities, and improving CBR-bank cooperation on bottom-up testing.
- Developments since previous FSAP:
  - CBR now conducts regular top-down single-factor tests and combined multi-risk tests; developed an econometric model-based top-down macro stress testing framework (one year horizon).
  - Bottom-up tests shared scenarios with top-down and included a very severe stress case.
- Suggested improvements:
  - Reduce number of banks for model efficiency (top 250 banks hold 95 percent of assets) to enable panel VAR or Arellano-Bond frameworks.
  - Extend horizon to 2–3 years (example: annual NPL change 2008–09 was 5.8 percentage points; two-year change to early 2010 was 7½ percentage points).
  - Consider a smaller robust macro model with satellite models to balance econometric robustness and model consistency.
  - Extend credit risk modeling toward Basel II-type portfolio loss concepts (credit VaR) as data permit; current CBR model assumes 100 percent provisioning for new NPLs without collateral mitigation.
  - Improve results presentation to highlight drivers and cross-sectional differences.

### Data gaps constraining analysis
- Supervisory data lack effective maturity dates for assets and liabilities, limiting liquidity stress tests.
- Data on liquid assets are too broad (mixing immediately available assets, assets available within 30 days, and the rest).
- Limited breakdown by underlying assets or securities (repo-ability vis-à-vis the central bank).
- Recommendation: fill data gaps promptly using existing reporting to enable granular systemic liquidity stress tests for central bank systemic liquidity management.

### Top-down testing methodology (high-level)
- Sample: bank-by-bank supervisory data as of end-2010 covering all banks (1012).
- Resilience benchmark: 10 percent CAR; metrics reported included losses in percent of capital, core capital ratio, liquidity ratio, and capital shortfalls to reach minimum CAR.
- Single-factor tests:
  - Credit shocks assume 100 percent provisioning for new NPLs and exclude collateral mitigation.
  - Interest rate valuation losses estimated using duration data; impact through interest income/expenses not incorporated.
  - Valuation losses for trading book and portion of AFS follow Basel II guidance.
  - Liquidity stress assumed immediate one-month extreme conditions with three asset liquidity categories and fire-sale discounts; no cash inflows except from fire sales.
  - Contagion: a bank “defaults” on interbank liabilities when total losses amount to 75 percent of capital in contagion stage.
- Macro scenario tests:
  - Econometric model linking macro variables and credit risk (corporate and household NPLs), one-year horizon. Capital buffers include existing capital and projected current year gross profits (assuming 0 payout).
  - Asset growth under severe stress about 2 percent; RWA change in line with net asset growth and loan repayments (downgrades do not affect RWA size).
  - Liquidity behavioral assumptions allow limited interbank market access at baseline + 1000 basis points and CBR collateralized lending with official repo-haircuts.

### Shock calibrations and scenario assumptions (preserved verbatim)
- Single-factor shocks: increases in NPL ratio by about 5 and 8 percentage points in two scenarios; simultaneous default of top five borrowers; liquidity withdrawals and fire sales; market shocks and interbank contagion.
- Macro scenarios:
  - Baseline: assuming about 4 percent annual GDP growth in 2011–12.
  - Pessimistic: assuming a 4½ percent points reduction in GDP growth relative to baseline.
  - Severe: assuming an 8 percentage points reduction relative to baseline (equivalent to 1.7 standard deviations using 2000–2010 data).
- Single-factor liquidity outflows and haircuts: household deposit— 20 percent; corporate settlement accounts—20 percent; corporate deposits—10 percent; cross-border interbank deposit—30 percent. Haircuts: high liquid asset—5 percent; liquid asset—20 percent; low-liquid assets—60 percent. No access to domestic interbank market in certain single-factor tests.
- Macro severe scenario selected parameters (preserved verbatim):
  - Real GDP growth rate, in percent: Baseline 4.0, Pessimistic -1.0, Severe -4.0
  - Oil price, USD/barrel: Baseline 70, Pessimistic 50, Severe 43
  - Inflation (CPI), in percent: Baseline 9.0, Pessimistic 11.0, Severe 7.3
  - Ruble depreciation (basket), in percent: Baseline 10, Pessimistic 20, Severe 26
  - Increase in interest rates of state securities (parallel shift), bps: Baseline 200, Pessimistic 300, Severe 350
  - Increase in interest rates of corporate securities (parallel shift), bps: Baseline 500, Pessimistic 900, Severe 1000

### Single-factor test highlights
- Credit shocks generally higher for smaller banks; corporate loan portfolio the main risk except for foreign-owned banks where household exposures are significant.
- Under type B credit shock (adding 1.65 standard deviation to latest actual NPL ratio), most bank types could experience losses near 40 percent of capital.
- Concentration risk: loans to largest five borrowers equal 10 percent of system loans and nearly 25 percent for smaller banks; loss could exceed half of capital for smaller and large private banks if top five borrowers default (LGD assumed 100 percent).
- Liquidity risk: second to credit risk; smaller regional banks and foreign-owned banks heavily affected due to deposit dependence and cross-border funding reliance; state-owned banks least affected.
- Large private banks: most vulnerable to liquidity shocks due to low highly liquid assets and weaker capital; only group failing one regulatory liquidity ratio in stress.
- Market risks: generally small relative to capital; interest rate shocks more important than equity shocks.

### Combined and contagion impacts
- Combined single-factor shocks (credit, liquidity, market): potential gross losses about one half of existing bank capital in memo figures; without gross profits buffer more than half of capital could be wiped out, failing over 300 banks and reducing system average capitalization below minimum requirement.
- Recapitalization cost after combined single-factor shocks: less than 2 percent of GDP.
- Interbank contagion on top of combined shock could add losses amounting to 13 percent of capital; contagion layering could imply additional failures (over 600 banks in extreme layering); recapitalization cost to recoup contagion losses: 1⅓ percent of GDP.

### Macro severe scenario results (selected headline figures)
- Headline CAR decline: 3.5 percentage points to 14.1 (system-wide CAR remains 14 percent in macro severe scenario with about 8 percent of the system below 10 percent CAR).
- Banks failing below 10 percent CAR: failing 75 banks (8 percent of the system).
- Total gross losses: about 35 percent of capital (mainly credit losses).
- A third of losses compensated by current year gross profit, leaving net losses of 22 percent of capital.
- Recapitalization cost to achieve 10 percent minimum regulatory capital requirement: 0.3 percent of GDP.

### Bottom-up versus top-down comparison (selected)
- Bottom-up headline CAR: 14.1 (slightly higher than top-down macro result for the same sample due to higher gross profits projected by banks).
- Projected losses similar: about 23 percent of capital in bottom-up and top-down macro severe scenario for the bottom-up sample.
- State-owned banks in bottom-up: projected more moderate losses and higher profits (about double the top-down profit projection).
- Private banks in bottom-up: projected substantially higher losses across risk types and more pessimistic profits than macro-model.

### Bottom-up severe scenario sample statistics (preserved verbatim)
- Bottom-up sample, actual data as of end 2010:
  - CAR (in percent): 17.6 19.1 16.1 14.1 (by group in table)
  - Core capital ratio (in percent): 9.8 10.0 11.4 7.8
  - NPL ratio (in percent): 9.3 9.7 10.9 6.1
- Bottom-up results, severe scenario:
  - CAR (in percent): 14.1 17.0 12.8 5.6
  - Total losses (in percent of capital): 23.2 16.5 26.2 47.8
  - Profit (in percent of capital): 13.6 19.2 6.1 -2.1
  - Changes in NPL ratio (in percentage points): 3.0 2.4 4.9 3.4

### Macro-financial model and NPL estimation (preserved equations and stats)
- Macroeconomic model: over 20 econometric equations covering household, real, external, monetary, and fiscal sectors; most estimation uses quarterly year-on-year change data for 1998–2010, using OLS.
- Household credit NPL ratio model (equation preserved verbatim):
  - Y = -1.5369 + 0.2359*X1[t] – 0.0353*X2[t] + 8.8125*X3[t-4] – 0.7167*X4[t-2] – 1.5575*X5[t]
  - Estimation highlights: R square 0.86; Adjusted R square 0.85; Durbin-Watson Statistic 1.30.
- Corporate credit NPL model (equation preserved verbatim):
  - Y = 11.7302 – 0.5807*X1[t] – 0.1227*X2 – 0.0927*X3[t-1] + 0.1052*X4[t] – 8.8787*X5[t] + 0.2026*X6[t-3]
  - Estimation highlights: R-Square 0.85; Adjusted R square 0.84; Durbin Watson Statistic 0.89.
- Interpretation:
  - For household credit, ruble depreciation affects credit quality negatively.
  - For corporate credit, ruble depreciation affects credit quality favorably due to corporate FX income; oil price is the most quantitatively important source of risks for corporate NPLs.

### Allocation of aggregate NPL changes to individual banks
- CBR allocates sector-wise NPL increases to individual banks using a coefficient based on:
  1. Institution’s current credit quality (relative to sector average);
  2. Size of the institution;
  3. Institution's propensity to credit risk (higher when NPL ratio more volatile than others).

*Source: Executive Summary (stress testing and financial stability assessment), IMF FSAP materials; Central Bank of the Russian Federation.*

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

### Executive Summary

### Stress Testing Framework and Scope
- The stress testing exercise for this FSAP is based on the existing Central Bank of Russia (CBR) approaches, as well as a separate bottom-up exercise.
- Top-down tests:
  - Use bank-by-bank supervisory data as of end-2010 and cover all existing banks in the system (1012) (implemented by the CBR).
  - Include single factor tests and macro-scenario tests.
- Bottom-up exercise:
  - Includes 15 top banks, covering about 57 percent of the system.
  - The CBR and the FSAP team agreed on stress scenarios and the same macroeconomic assumptions as the top-down exercises.
- Resilience benchmark:
  - Assessed using the current minimum regulatory capital adequacy ratio (CAR) requirement of 10 percent.
- Risk coverage:
  - Single factor tests examined instantaneous impact of credit, concentration, market (foreign exchange, equity, and interest rate), liquidity, and interbank contagion risks.
  - Macro scenario test focused on credit, market and liquidity risks with one year risk horizon.
  - The severe scenario examined the impact of a macroeconomic shock equivalent to 1.7 standard deviation shock on real GDP growth rate.

### Key Stress Test Results
- Overall resilience:
  - The results suggest that the Russian banking system is, on the whole, resilient to a variety of macroeconomic and financial shocks.
- Capital and losses:
  - Gross losses to the banking sector in a tail economic event might be substantial (about 35 percent of capital), mainly owing to credit losses.
  - About one-third of these losses would be offset by profits.
  - Although the system as a whole would maintain a 14 percent capital ratio, banks representing about 8 percent of the system (mainly large private banks) would fall below the 10 percent minimum capital ratio.
- Liquidity risk:
  - Acute, systemic liquidity shocks (including on foreign exchange liquidity) may burden the banking sector significantly in a very short period of time.
  - For most Russian banks, funding would be very volatile during a stress, including individual deposits, which are usually more stable in other countries.
  - The stress tests do not take into account the policy reaction of the CBR in the event of a liquidity shock.
- Market and FX risk:
  - Valuation losses on securities, especially bonds, could be notable, reflecting the recent increase in securities investment.
  - Direct foreign exchange valuation risk is negligible, given the small open foreign exchange position.
- Concentration risk:
  - Concentration risks are significant, especially for smaller banks, given the high degree of credit concentration in the Russian banking system.
- Bottom-up vs top-down:
  - Bottom-up test yielded broadly similar results to the top-down tests.

### Structural and Supervisory Weaknesses (Adjustments and Impact)
- Adjustments for vulnerabilities:
  - Structural and supervisory weaknesses imply the system may be more fragile and vulnerable than the stress tests suggest.
  - The impact of adjusting the baseline balance sheet and capital position for these vulnerabilities dwarfs the effects from most economic shocks.
- Specific adjustment effects:
  - Adjusting for the potential overestimation of loan quality and for low provisions could give much larger shocks to capital than the aforementioned economic shocks.
  - Adjusting for provisions as described could wipe out as much as one-third of the capital.
  - Adjusting for the credit quality of extended maturity loans can reduce capital by over 20 percent, especially for state-owned and large private banks, which tend to keep a larger share of restructured loans in standard category.
- Data caveat:
  - In the absence of hard data, the magnitude of these adjustments is driven to some extent by subjective assumptions.

### Mitigating Factors and Policy Capacity
- Capital and profitability:
  - High capital buffers and strong profitability could function as shock absorbers.
- Fiscal and external capacity:
  - Estimated economic cost of recapitalization is small and manageable, given:
    - The relatively small size of the Russian banking sector relative to GDP.
    - Low government debt.
    - High reserves.
- Policy precedent:
  - Government interventions to support and recapitalize weak banks may increase moral hazard, but the ability and availability of resources to intervene decisively—as in the recent crisis—mitigates systemic risk.

### Financial Sector Structure and Key Issues
- Size and composition:
  - As of end-2010, total assets of financial institutions were around 80 percent of GDP.
  - Banks accounted for over 90 percent of the total assets.
  - Bank deposits are just above 40 percent of GDP.
- Number and concentration:
  - As of end-2010, there were 1,012 banks operating in Russia, a decline by 46 from January 1, 2010, and about 250 less than five years earlier.
  - The CBR’s minimum capital requirement for banks was raised to Rub 90 million in January 2010, with further increases to Rub 180 million in 2012 and Rub 300 million in 2015 planned.
  - At the end of 2010, the share of the state-owned banks in total deposits was 52 percent.
- Balance sheet structure:
  - Majorities of the assets are loans (mostly to industries), followed by securities (mostly in government and corporate bonds, followed by equities) and interbank transactions (60 percent are vis-à-vis non-resident banks).
  - Borrowings from non-resident declined to 13 percents of the book at end-2010, down from 20 percent in 2007, and they are mostly long-term.
  - Liquidity support from the CBR, which rose substantially in 2008, was largely withdrawn by end-2010.
- Structural and supervisory concerns:
  - Insufficiently diversified client base and high concentration:
    - Loans to top five borrowers, on average, amount to 5 percent of the assets and 50 percent of capital.
    - Operations of many small banks are concentrated to their owners or affiliated parties.
    - Regulatory deficiency: narrow definition of related parties may understate true concentration.
  - Deposit volatility:
    - Deposits are highly volatile for the majority of the banks; in stress, depositors tend to shift funds to large state-owned banks.
  - Weaknesses in loan-loss provisions:
    - Reported provisions are on the lower end of the ranges determined in the regulations, even when collateral is taken into account.
    - The quality of collateral varies widely; costs of seizing collateral and ability to sell it in distressed conditions may not be accurately reflected.

### Impact of the Crisis and Recent Performance
- Two waves of crisis impact:
  - First wave (late 2008) – liquidity:
    - Substantial deposit withdrawals, reaching almost 20 percent of total even in some large banks in a month, although loss of deposits was considerably smaller for the system as a whole.
    - Funds from non-residents were withdrawn sharply, losing $70 billion (about 7 percent of total assets) between September 2008 and end 2009.
    - Liquidity pressures spread through the domestic interbank market, contributing to a brief but sharp spike in the interbank interest rate in January 2009.
  - Second wave (2009) – credit:
    - The NPL ratio jumped by about 7 percentage points between end-2007 and early 2010.
    - Provisioning costs weakened bank profitability substantially, although the sector overall continued to make net profits even in the middle of the crisis.
- Policy response and outcomes:
  - Ruble stability:
    - The CBR used sizable reserves (nearly $600 billion as of mid-2008) to support a gradual and predictable depreciation of the ruble.
    - Total reserve loss between August 2008 and January 2009 amounted to over $200 billion.
  - Emergency liquidity support:
    - Lending from the CBR amounted to about 12 percent of bank assets at end 2008.
  - Capital injections and deposit insurance:
    - Capital injections to several government-owned banks amounted to Rub 505 billion (1.3 percent of GDP).
    - Additional subordinated loans from VEB or the CBR totaled Rub 904 billion (2.2 percent of GDP).
    - Deposit-insurance limit was raised and the deposit insurance agency was allocated additional resources and powers.
  - Temporary regulatory forbearance:
    - Loan delinquency thresholds were relaxed (corporate loan overdue defined as 30 days, up from 5 days; retail 60 days, up from 30 days).
    - Restructured loans were allowed to remain in their original classification category.
    - These steps are estimated to have saved banks Rub 300 billion in provisions (7 percent of capital) in mid-2009; by late 2010 the estimated savings had declined to Rub 80 billion (2 percent of capital).
    - All forbearance measures were withdrawn as of July 2010, although grandfathering effects are expected to remain for another year.

*Source: Executive Summary (stress testing and financial stability assessment), IMF FSAP materials.*

### 8.      After the crisis, the performance of banks started to recover. Partly as a result of

### _cr11334 - 8.      After the crisis, the performance of banks started to recover. Partly as a result of

### Post-crisis banking performance
- Aggregate capital adequacy ratio stood at 18.1 percent in December 2010, well above the prudential minimum of 10 percent.
- The NPL ratio has stabilized despite the termination of regulatory forbearance measures.
- Funding conditions improved: household and corporate deposits grew strongly, allowing the CBR to discontinue its emergency liquidity support.
- Bank assets are growing again, but at a much slower rate than before the crisis and with growth more towards securities and less to credits.
- Bank profitability rebounded in 2010, largely reflecting lower provisioning costs; profitability in 2010 was still weak by Russian standards, but it was higher than in comparator countries.

### Performance differences across bank groups
- State-owned banks:
  - 20 state-owned banks hold 46 percent of the system’s assets.
  - Well capitalized, but loan quality is weaker than that of other banks.
  - Relatively cheap and stable household deposits and quick access to CBR refinancing allow them to hold less excess liquidity.
- Foreign-owned banks:
  - 108 banks, 19 percent of the system’s assets.
  - Well capitalized, typically lack branch networks and rely substantially on external funding (particularly from their parent banks).
  - Household loans represent the largest share of their credit portfolios.
- Large private banks:
  - Relatively low capitalization and profitability, but also a relatively low share of nonperforming loans.
- Small private banks:
  - Around 700 smaller banks have aggregate capital and liquidity ratios well above the system’s average.
  - Face difficulties accessing the interbank market and lack big foreign parents.
  - Higher concentration risks on both asset and liability sides and report weaker profitability.

### Hidden vulnerabilities and sources of overstated credit quality
- Foreclosed assets: some assets (especially those collateralized by real estate) that are not earning any cash are said to be reported at an overvalued price on balance sheet.
- Opaque transfer of distressed assets: practice of transferring distressed assets to off-balance sheet entities (special purpose vehicles or distressed asset funds) that are often not covered under consolidated reports.
- Restructured loans:
  - The share of restructured loans among large loans increased visibly at the height of the crisis and remained at a higher level than the pre-crisis time.
  - Loans with extended maturities are perhaps more likely to have lower credit quality.
  - More than 90 percent of the restructured loans are in the performing categories.
  - There are no visible changes with this figure so far after the withdrawal of the forbearance measure that allowed banks to avoid classifying the restructured loans to lower quality categories.

### Stress testing: framework, practice, and evolution
- Historical FSAP (2008) coverage:
  - Top-down single factor tests and bottom-up tests were used, with methodological and data gaps in macro stress testing.
  - Tests included: (i) top-down single factor tests (for all living banks) on credit, market, liquidity risks using supervisory data; and (ii) bottom-up test (for five major banks) on the same set of risks using banks’ internal data.
  - Regulatory capital benchmark in tests: 10 percent minimum capital to risk-weighted-assets ratio.
  - Major FSAP recommendations included: (i) developing macro scenario stress testing framework; (ii) consider implementing credit risk VaR models; (iii) collect data on effective maturities for liquidity gap analysis; (iv) improve cooperation between the CBR and banks regarding bottom-up stress testing.
- Developments since previous FSAP:
  - The CBR now conducts regular top-down single factor tests and a combined test of multiple risk factors (credit, liquidity, market) that mimics a macroeconomic scenario.
  - A recently developed econometric model-based framework conducts top-down macro stress tests linking risk factors to macroeconomic indicators (model detailed in the appendix).
  - The macro stress testing has a one year horizon and includes credit, market, and liquidity risks.
  - Bottom-up tests in this FSAP shared stress scenarios with top-down macro scenarios and included a very severe stress case.
- Current FSAP testing scope and approach:
  - Top-down single-factor tests clarify which risks are relatively most relevant with limited emphasis on adequacy of capital levels.
  - Shocks to individual risk factors are calibrated based on historical developments and expert judgments.
  - Macro-scenario tests assess adequacy of capital buffers in extreme but plausible tail events; shocks are correlated systemic shocks driven by common macroeconomic variables.
  - Bottom-up tests use banks’ internal data and models to cross-check top-down results and incorporate risk-augmenting or mitigating positions not reflected in supervisory data.

### Strengths of the macro stress testing framework
- Comprehensive coverage of the banking sector: the test covers all the banks in the system.
- Comprehensive coverage of risk types: covers credit risk, market (equity, exchange rate, interest rate) risks, and liquidity risks.
- Attempts to include various macro-financial linkages: model consists of over 20 econometric equations covering the real, external, fiscal, and financial sectors, including second-round feedback effects.
- Estimated core macro-financial linkages (e.g., relationship between NPL ratio and macroeconomic factors) appear in line with the actual experience in 2009 and earlier IMF estimates.

### Potential areas for improvement and recommendations
- Reduce the number of banks included for efficiency and more robust econometric performance:
  - Out of over 1000 banks in the model, the top 250 banks (including all state-owned, foreign-owned, and large private banks) have 95 percent of the sector’s assets.
  - Limiting the sample could allow use of panel VAR or Arellano-Bond type micro-econometric frameworks; smaller banks’ resilience could be assessed by single factor tests.
- Extend stress testing time horizon to 2-3 years and beyond:
  - Example: annual change in NPL ratio between end 2008 and 2009 was 5.8 percentage points; change for about two years between end 2007 and early 2010 when NPL ratio peaked at 10 percent was 7½ percentage points.
  - Given the existing macro model structure, such an extension should be fairly straightforward.
- Weigh benefits of expanding the macro model against maintaining theoretical and econometric robustness:
  - Current strategy constructs a large system of equations estimated by OLS including year-on-year changes, which can be susceptible to omitted variables and auto-correlated errors.
  - Alternative: smaller main macro model with separate satellite models to incorporate advanced techniques and theories, acknowledging potential trade-offs in overall model consistency.
- Extend credit risk modeling toward Basel II (and beyond)-type portfolio loss concepts (such as credit VaR) as data permit:
  - Current CBR model projects credit loss by estimating increases in NPL ratios for household and corporate loans and assumes 100 percent provisioning for new NPLs (without risk mitigation using collaterals).
  - Reliable PD and LGD data are limited; the credit registry is fairly new and fragmented.
  - Some corporate borrower data exist at the CBR and increasing private sector estimates; pilot utilization of existing data is encouraged.
- Improve results presentation to highlight driving factors and cross-sectional differences:
  - Consider incorporating charts or tables showing (i) key indicators of scenario severity (e.g., NPL increases, liability run); (ii) summary statistics by bank type (state-owned, private, etc.); and (iii) broader cross-sectional distribution.
  - Presentation in this FSAP can serve as a starting point.

### Data gaps constraining analysis
- Supervisory data lack effective maturity dates for assets and liabilities, limiting liquidity stress tests.
- Data on liquid assets are too broad (mixing immediately available assets, assets available within 30 days, and the rest).
- Limited breakdown by underlying assets or securities (repo-ability vis-à-vis the central bank).
- Recommendation: fill data gaps promptly, initially by utilizing existing reporting, to enable systemic liquidity stress tests with granular data for central bank systemic liquidity management.

*Source: IMF staff discussion in the provided chapter excerpt.*

### 16.      The two top-down tests and bottom-up test share broadly similar simulation

### _cr11334 - 16.      The two top-down tests and bottom-up test share broadly similar simulation

### Simulation scope and high-level methodology
- Top-down single-factor and macro scenario tests use bank-by-bank supervisory data as of end-2010 and cover all existing banks in the system (1012).
- Bottom-up exercise covers 15 major banks, covering over 55 percent of the system by assets.
- The CBR’s Supervision Department coordinated the bottom-up test using the same macroeconomic assumptions as the top-down exercises.
- Resilience benchmark: current minimum regulatory capital adequacy ratio (CAR) requirement of 10 percent.
- Other reported metrics: losses in percent of capital, core capital ratio, liquidity ratio, and capital shortfalls necessary to achieve the minimum CAR.

### Top-down single-factor tests — risk coverage and calculation rules
- Tested risk factors:
  - Credit risks (including adjustment for adequacy of provisioning requirements, recent forbearance, and the quality of restructured loans; increases in NPL ratio for household and corporate loans; concentration risk — default of top five borrowers).
  - Market risks (interest and exchange rate and equity prices).
  - Liquidity risks.
  - Contagion risks within the banking system.
- Key methodological characteristics:
  - Risk-weighted-assets (RWA) are kept constant and no current year profit is included in capital.
  - Credit shocks: 100 percent provisioning rate assumed and all collaterals excluded for new NPLs.
  - Interest rate risks: valuation losses from bonds estimated using duration data; impact through interest income/expenses is not incorporated.
  - Valuation losses from interest rate and equity price changes assessed for securities held in trading book and a portion of available-for-sales (AFS) accounts, following Basel II guidance.
  - Only direct impact of exchange rate fluctuation assessed, using net open foreign exchange positions.
- Liquidity stress (acute, immediate-one month):
  - Assumed extreme conditions (complete shut-down of interbank market, severe stressed haircut severer than the CBR’s repo haircut for Lombard list securities) and no cash inflows except from fire sales.
  - Three asset liquidity categories: highly liquid (available within a day), liquid (available within a month), and illiquid (others); fire-sale discounts applied.
  - Withdrawal sources examined: household deposits; funds in settlements, current, and other accounts of non-financial organizations; deposits by non-financial organizations; cross-border interbank deposits.
  - Lack of liquid asset data by currency prevents liquidity tests by currency.
- Combined test: CBR regularly tests combined shocks (increases of NPL ratios for corporate and household sectors, ruble depreciation, equity price declines, interest rate increases, and liquidity withdrawals).
- Contagion risk: assessed using matrix of interbank positions; in contagion stage a bank “defaults” on liabilities held by other banks when total losses amount to 75 percent of the capital.

### Top-down macro scenario tests — design and assumptions
- An econometric model linking macro variables and credit risk (headline NPL ratio for corporate and household sectors separately) estimated by Prognoz; assumptions on market risks and liquidity withdrawal rates also derived from the macro model.
- Time horizon: One year.
- Risks measured: credit (corporate and household loans), market (equity, exchange rate, interest rate), and liquidity. Equity and interest valuation shocks applied to trading and portion of AFS securities; new NPLs require 100 percent provisioning and no collateral mitigation.
- Contagion risk assessed as an add-on exercise.
- Capital buffers: include existing capital and all projected current year gross profits, assuming 0 payout ratios. In stressed scenario banks make losses and distribute nothing.
- Gross profits projected as a function of interest rate assumptions; interest incomes from performing customer loans assume constant margin; interest income/expenses from interbank loans assessed at a stressed rate of baseline + 1000 basis points; commissions and fee income projected in line with trend growth.
- Asset growth: macro model generates bank asset growth; under severe stress growth is moderate at about 2 percent. RWA changes in line with net asset growth and loan repayments; downgrade/increases in PD and LGD do not affect size of RWA.
- Liquidity behavioral assumptions (one-year horizon, moderate conditions):
  - Access to interbank market allowed at baseline + 1000 basis points, limited to counterparties with prior transactions.
  - Access to CBR collateralized lending allowed using official repo-haircuts to Lombard list securities.
  - Cash inflows from performing loan repayments included, mitigating shortages.

### Bottom-up tests — approach and alignment with top-down
- Bottom-up assessed combined credit, market (equity prices, interest rate, exchange rate), and liquidity risks; impact measured by regulatory capital, core capital ratio, and losses in percent of existing capital.
- Banks given three methodological choices:
  - (A) take broad macroeconomic assumptions given by the CBR and use their internal macro model to translate to risk factors;
  - (B) take combined single-factor assumptions given by the CBR and apply to balance sheet data;
  - (C) report own stress testing methodology and results.
- All participating banks chose either (A) or (B).
- CBR did not impose a specific method to project gross profits.
- Bottom-up and top-down exercises shared the same macroeconomic and combined single-factor assumptions.

### Assumptions and shock calibrations
- Single-factor shocks calibrated broadly in line with 2009 experience and previous FSAP, including increases in NPL ratio (by about 5 and 8 percentage points in two separate scenarios), simultaneous default of top five borrowers, liquidity shock (withdrawal of liabilities followed by fire sales), market risks, and interbank contagion risks.
- Macro scenarios:
  - Baseline: assuming about 4 percent annual GDP growth in 2011–12.
  - Pessimistic: assuming a 4½ percent points reduction in GDP growth relative to baseline.
  - Severe: assuming an 8 percentage points reduction relative to baseline.
  - Growth shocks in pessimistic and severe scenarios equivalent to 1 and 1.7 standard deviations of GDP growth using data for 2000–2010.
- Single-factor liquidity risk parameters (withdrawal rates and haircuts) calibrated on historical episodes and expert judgment; cross-border interbank loan shock incorporated.
- In macro stress tests, discounts for liquid assets based on CBR repo haircut.
- Differential liquidity stress magnitudes applied across bank types (state-owned banks subject to more moderate deposit withdrawal rates than others).
- Baseline adjustments to reflect regulatory forbearance and structural/supervisory weaknesses:
  - (i) estimated impact of forbearance at end-2010 added to provisions (adjustment roughly 2 percent of capital based on proxy).
  - (ii) restructured loans assumed fully provisioned.
  - (iii) provisions in each loan category raised to midpoint of regulatory range and poor quality collateral assumed to have no value.
- Note: these adjustments are ad hoc and extreme but intended to gauge underlying portfolio strength.

### Results — adjustments and sensitivity
- Adjustment for forbearance effects on loans with grandfathered effects: minor, about 2 percent of capital; seven banks become undercapitalized but their system share is negligible.
- Adjustment for credit quality of restructured loans or raising provisioning rate to midpoint could amount to 20–35 percent of capital, potentially impacting 61–251 banks (9–38 percent of the system). Effects vary by bank ownership type.

### Results — single-factor tests (highlights)
- Credit shocks generally higher for smaller banks, reflecting higher volatility of historical NPL ratios; most risks stem from corporate loan portfolio except for foreign-owned banks where household exposures are significant.
- Under type B credit shock (adding 1.65 standard deviation to latest actual NPL ratio), most bank types could experience losses near 40 percent of capital.
- Small and medium-sized banks in Moscow region exhibit stronger resilience despite higher credit shocks due to stronger capital positions.
- Concentration risk: loans to largest five borrowers equal 10 percent of system loans and nearly 25 percent for smaller banks; more than half of capital could be lost for smaller and large private banks if top five borrowers default (LGD assumed 100 percent).
- Liquidity risk: second to credit risk; smaller regional banks most affected by liability withdrawals due to dependence on household and corporate deposits; foreign-owned banks heavily affected due to reliance on cross-border interbank funding; state-owned banks affected least due to explicit/implicit guarantees and deposit shifts.
- Large private banks are the most vulnerable group to liquidity shocks due to low amounts of highly liquid assets and weaker capital; this group is the only one failing one regulatory liquidity ratio in the stress scenario.
- Market risk losses generally small relative to capital; ruble fluctuation impact negligible due to near 0 net open FX positions; interest rate shocks more important than equity shocks (¾ of securities are bonds).
- Combined single-factor shocks (credit, liquidity, market): potential losses could amount to about a half of existing bank capital. Without current year gross profits buffer, impact could wipe out more than half of capital, fail over 300 banks, and reduce system average capitalization below minimum requirement.
- Recapitalization cost to achieve 10 percent minimum requirement after combined single-factor shocks: less than 2 percent of GDP.
- Interbank contagion effects (upon combined shock) could add losses amounting to 13 percent of capital; layered on top of severe shocks, contagion analysis could exaggerate marginal failures (failing additional 600 plus banks). Recapitalization costs to recoup contagion losses: 1⅓ percent of GDP.
- Across groups, small and medium-sized regional banks most affected by contagion due to high domestic interbank exposures as share of assets.

### Macro scenario test results (selected)
- In the severe scenario, the system-wise CAR declines by 4 percentage points to

*Source: IMF FSAP chapter content as provided in the source PDF.*

### 14.1 percent, failing 75 banks (8 percent of the system, Figure 2). The recapitalization

### _cr11334 - 14.1 percent, failing 75 banks (8 percent of the system, Figure 2). The recapitalization

### Stress test headline results
- Headline CAR declines by 3.5 percentage points to 14.1 for the whole sample of participated banks.
- The system-wide regulatory capital ratio (CAR) remains 14 percent in the macro severe scenario, but banks representing about 8 percent of the system (mainly large private banks) would fall below the 10 percent minimum capital ratio.
- Total gross losses to the banking sector in a tail economic event are about 35 percent of capital, mainly owing to credit losses.
- A third of the losses are compensated by current year gross profit, leaving net losses of 22 percent of capital.
- The recapitalization cost to achieve the 10 percent minimum regulatory capital requirement is small (0.3 percent of GDP).

### Bottom-up versus top-down findings
- Bottom-up stress test results are broadly in line with the top-down macro scenario test (severe scenario).
- Differences:
  - Bottom-up headline CAR (14.1) is slightly higher than top-down macro result for the same sample due to higher gross profits projected by banks in the bottom-up exercise.
  - Projected losses are fairly close between top-down macro scenario test and bottom-up tests (about 23 percent of capital).
  - State-owned banks projected much more moderate losses and higher profits in bottom-up tests (about double the top-down macro scenario test profit projection).
  - Private Russian banks projected substantially higher losses across all risk types and more pessimistic profits than the macro scenario test.
- Reasoning:
  - Private banks often used approach B (combined single-factor test), assigning severer shocks to each risk factor than macro-models, producing losses close to top-down combined single-factor tests (excluding liquidity).

### Risk composition and bank-group vulnerabilities
- Credit risks are the most important source of losses, followed by valuation losses from securities.
- Liquidity shocks can be acute and systemic in a very short period (unlike credit losses which materialize over a year or two).
- Direct foreign exchange valuation risk is negligible due to small open FX positions; indirect FX risk (via credit risk) is material for households borrowing in FX without FX income.
- Concentration risks are significant, especially for smaller banks, given high credit concentration.
- Large private banks are the most vulnerable group across tests and shocks, primarily due to weak capital and liquidity positions.
- Foreign and small and medium sized banks face more severe liquidity pressures; smaller banks also suffer more severe credit shocks, but relative to buffers, large private banks are most impacted.

### Adjustments, vulnerabilities, and sensitivity to assumptions
- Structural and supervisory weaknesses imply the system may be more fragile than stress tests suggest.
- Adjusting the baseline balance sheet and capital position for vulnerabilities produces larger impacts than most economic shocks:
  - Adjusting for provisions and potential overestimation of loan quality could wipe out as much as one-third of capital.
  - Adjusting for credit quality of extended maturity loans can reduce capital by over 20 percent, especially for state-owned and large private banks that keep larger shares of restructured loans in the standard category.
- The magnitude of these adjustments is partly driven by subjective assumptions in the absence of hard data.

### Mitigating factors and fiscal implications
- High capital buffers and strong profitability can function as shock absorbers.
- Even when capital injections are required, the estimated economic cost of recapitalization is small and manageable given:
  - Relatively small size of the Russian banking sector relative to GDP,
  - Low government debt,
  - High reserves.
- Government interventions to support and recapitalize weak banks may increase moral hazard, but the ability and availability of resources to intervene decisively mitigates systemic risk.

### Selected exact figures and scenario parameters (preserved verbatim)
- Headline CAR decline: 3.5 percentage points to 14.1
- Banks failing below 10 percent CAR: failing 75 banks (8 percent of the system)
- Recapitalization cost to regain 10% CAR: 0.3 percent of GDP
- Total losses from various sources of risks: 34 percent of capital
- Net losses after gross profit: 22 percent of capital
- Gross profits offsetting losses: about 11 percent of capital
- Combined single-factor gross losses (memo): Total losses 51 42 49 69 34 45 (table figures by group)
- Macro severe scenario: Regulatory capital ratio (CAR) 14.1 14.9 15.3 10.2 22.2 16.7 (table figures by group)
- Net losses, % capital (severe scenario): -22 -20 -22 -35 -17 -25
- Total losses, % capital (severe scenario): -34 -32 -33 -48 -27 -33
- Capital shortfall % GDP (to regain 10% CAR, severe): 0.3 0.1 0.0 0.1 0.0 0.0
- Macro scenario assumptions (selected):
  - Real GDP growth rate, in percent: Baseline 4.0, Pessimistic -1.0, Severe -4.0
  - Oil price, USD/barrel: Baseline 70, Pessimistic 50, Severe 43
  - Inflation (CPI), in percent: Baseline 9.0, Pessimistic 11.0, Severe 7.3
  - Ruble depreciation (basket), in percent: Baseline 10, Pessimistic 20, Severe 26
  - Increase in interest rates of state securities (parallel shift), bps: Baseline 200, Pessimistic 300, Severe 350
  - Increase in interest rates of corporate securities (parallel shift), bps: Baseline 500, Pessimistic 900, Severe 1000
- Single-factor shock assumptions (excerpt):
  - Credit risk: (A) 1.65 stdev shock on historical average; (B) 1.65 stdev shock on actual.
  - Market risks: FX— 20 percent depreciation; equity— 30 percent decline; interest rate— +300 (900) bps for government (corporate) bonds.
  - Liquidity risk outflows: household deposit— 20 percent; corporate settlement accounts—20 percent; corporate deposits—10 percent; cross-border interbank deposit—30 percent. Haircuts: high liquid asset—5 percent; liquid asset—20 percent; low-liquid assets—60 percent. No access to domestic interbank market.

*Source: Central Bank of the Russian Federation and IMF staff calculation.*

### 1.65 stdev shock on actual

### 1.65 stdev shock on actual

### Key post-shock assumptions and scenario design
- Impact from losses from forbearance measures temporarily introduced during the crisis time up to July 2010.
- Loans whose maturity was lengthened default lose 100% of their values. At end 2010, total restructured loan amounted to 30% among large loans, a half of which were extended maturity loans.
- Raise provisioning rates of each loan category to the middle value of the regulatory acceptable ranges, including high quality collaterals only.
- Losses from asset firesales upon liability side shocks:
  - Withdrawal rates are 30 % for interbank liabilities from non-resident banks; 20% for individual deposits and funds from settlement accounts from corporations; 10% for deposits from corporations.
  - 5, 20, and 60 percent haircuts with highly liquid, liquid, and low-liquid assets respectively.
  - No access to domestic interbank market, including the CBR.
- Parallel shift up of bond yield curves.
- In addition to the combined shocks, including credit (A: historical distribution) liquidity and market shocks. In the contagion stage, a bank "defaults" 100% on its interbank borrowings when its losses from interbank contagion effects reach 75% of its stressed capital.
- Assuming, a 8 p.p. drop in GDP growth rate (i.e. 1.68 standard deviation using 2000-2010 data) from baseline.
- Loans to top 5 borrowers are about 50% of capital. Compared to other shocks, this scenario should have extremely lower probability to occur.

### Summary of bottom-up stress testing results (key statistics)
- Sample:
  - Number of banks 15285
  - Share in total banking sector by assets (in percent) 5 7 3 6 10 11 (as presented in table with grouping All State owned Foreign Other Sample)
- Test Methodology:
  - Approach A: Number of banks using internal macro model 6 14 1
  - Approach B: Number of banks using combined sensitivity tests 9 14 4
- Actual data as of end 2010, for bottom-up sample:
  - CAR (in percent) 17.6 19.1 16.1 14.1
  - Core capital ratio (in percent) 9.8 10.0 11.4 7.8
  - NPL ratio (in percent) 9.3 9.7 10.9 6.1
- Bottom-up results, severe scenario:
  - CAR (in percent) 14.1 17.0 12.8 5.6
  - Total losses (in percent of capital) 23.2 16.5 26.2 47.8
  - Profit (in percent of capital) 13.6 19.2 6.1 -2.1
  - Changes in NPL ratio (in percentage points) 3.0 2.4 4.9 3.4
- CBR's combined single factor test results, for bottom-up sample:
  - CAR (in percent) 8.5 10.6 7.0 3.6
  - Total losses (in percent of capital) 51.5 44.8 56.3 74.6
  - Changes in NPL ratio (in percentage points) 5.5 6.1 3.7 5.1
- CBR's macro stress test (severe scenario), for bottom-up sample:
  - CAR (in percent) 13.0 15.0 12.5 9.5
  - Total losses (in percent of capital) 23.1 21.9 21.6 27.9
  - Profit (in percent of capital) 11.2 10.9 11.9 11.5
  - Changes in NPL ratio (in percentage points) 3.6 3.4 4.1 4.2

### Top-down and combined test findings (high-level)
- Cumulative Distribution of Regulatory Capital Ratio shows material left-tail erosion under the severe macro test relative to Actual, end-2010 (figures present distributions by number of banks and by market share).
- Various adjustments (Forbearance (FB); Extended maturity + FB; Mid prov. 50% good collateral + FB) shift distributions noticeably toward lower CAR buckets.
- Credit risk: Changes in NPL ratio in stress tests:
  - NPL increase A: 5 percentiles of historical dist.
  - NPL increase B: actual + 1.65 stdev
- Combined single factor test: details of impact in percent of capital include components labeled Bond loss, Equity loss, FX loss, Fire sales loss, Credit loss, Contagion effects.
- Liquidity stress tests:
  - Withdrawal of liabilities in percent of assets (Deposit, individual; Settlement account, organizations; Deposit, organizations; Non-resident interbank liabilities) show losses by bank type.
  - Losses from asset fire sales in percent of capital presented by bank type.
  - Liquidity Ratio: Liquid assets/Short-term liabilities in percent — Actual, Single factor liquidity stress test, Regulation comparisons.
  - Liquidity Ratio: High liquid assets/demand liability in percent — Actual, Single factor liquidity stress test.

### Bottom-up detailed loss composition (by bank type)
- Bottom-up test results, losses and profits in percent of capital, by types of banks present contributions from:
  - Profit
  - Bond valuation loss
  - Equity valuation loss
  - FX loss
  - Liquidity loss
  - Credit loss
- Top-down macro scenario test results for bottom-up sample, losses and profits in percent of capital, by types of banks, show similar decomposition.
- Combined single factor top-down test result for bottom-up sample, losses and profits in percent of capital, by types of banks, list Bond valuation loss, Equity valuation loss, FX loss, Liquidity loss, Credit loss.

### Macroeconomic model with the banking sector (model structure and estimation)
- The macroeconomic model:
  - Consists of a set of over 20 econometric equations, covering the household, real, external, monetary, and fiscal sectors.
  - Most estimation uses quarterly year on year change data for the period of 1998–2010, using OLS.
- Core macro-financial link:
  - Links aggregate NPL ratios for corporate sector loans and household sector loans to macroeconomic variables.
  - Models estimated with data from 2000 to 2010, using monthly year-on-year growth rate of each variable.

Household credit NPL ratio model:
- Equation:
  - Y = -1.5369 + 0.2359*X1[t] – 0.0353*X2[t] + 8.8125*X3[t-4] – 0.7167*X4[t-2] – 1.5575*X5[t]
  - Y – Percentage point change in NPL ratio (the share of category 4, 5 loans) for credits to individuals
- Estimation results (Table 8):
  - A0 (constant) -1.5369, Standard error 2.4268, T-Statistic -0.6333
  - X1 – Change in consumer price index, percentage points. 0.2359, Standard error 0.0359, T-Statistic 6.5770
  - X2 – Rate of growth of public real disposable income, in percent -0.0353, Standard error 0.0170, T-Statistic -2.0813
  - X3 – Rate of growth of the exchange rate of the dollar against the ruble, in percent 8.8125, Standard error 1.3855, T-Statistic 6.3603
  - X4 – Change in refinancing rate, percentage points. -0.7167, Standard error 0.0935, T-Statistic -7.6659
  - X5 – Rate of growth of cash (М0), in percent -1.5575, Standard error 0.4421, T-Statistic -3.5234
  - R square 0.86
  - Adjusted R square 0.85
  - F statistic – p value 0.00
  - Durbin-Watson Statistic 1.30

- Model performance:
  - The chosen model produces functioning estimates of NPL, keeping track of the actual development, including during the 2008–2009 crisis time, fairly well.
  - For household credit, the ruble depreciation affects credit quality negatively.

Corporate credit NPL model:
- Equation:
  - Y = 11.7302 – 0.5807*X1[t] – 0.1227*X2 – 0.0927*X3[t-1] + 0.1052*X4[t] – 8.8787*X5[t] + 0.2026*X6[t-3]
  - Y – Percentage point changes with NPL ratio (share of category 4 and 5 loans) for credit to legal entities
- Estimation results (Table 9):
  - A0 (constant) 11.7302, Standard error 5.1076, t-statistic 2.2966
  - X1 – Rate of growth of the exchange rate of the dollar against the ruble, in percent -0.5807, Standard error 0.0552, t-statistic -10.5277
  - X2 – Rate of growth of production (GDP), in percent -0.1227, Standard error 0.0585, t-statistic -2.0982
  - X3 – Rate of growth of investment in fixed capital, in percent -0.0927, Standard error 0.0168, t-statistic -5.5217
  - X4 – Rate of growth of consumer price index, in percent 0.1052, Standard error 0.0139, t-statistic 7.5474
  - X5 – Change of price of oil, dollars per barrel -8.8787, Standard error 0.9724, t-statistic -9.1303
  - X6 – Change in refinancing rate, percentage points. 0.2026, Standard error 0.0160, t-statistic 12.6262
  - R-Square 0.85
  - Adjusted R square 0.84
  - F statistic, p-value 0.00
  - Durbin Watson Statistic 0.89

- Model performance and interpretation:
  - The chosen model produces functioning estimates of NPL, keeping track of the actual development, including during the 2008–2009 crisis time, fairly well.
  - For corporate credit, the ruble depreciation affects credit quality favorably, as the overall corporate sector have foreign exchange income to benefit from ruble depreciation.
  - In terms of the sources of risks, the model indicates that oil price is the most quantitatively important source of risks.

### Allocation of aggregate NPL changes to individual banks
- Because of the substantial number of banks in the system (over 1000), the CBR:
  - First constructed a model that relates aggregate NPL ratios to macroeconomic variables.
  - Then “allocated” the macro trends to individual banks’ NPL ratio.
- Allocation rule: The increase in sector-wise non-performing loan is allocated according to a coefficient, which is a function of:
  1. The institution’s current credit quality (relative to sector average);
  2. The size of the institution; and
  3. The institution's propensity to credit risk (which increases when the NPL ratio of the institution is more volatile than the others).

*Source: The Central Bank of Russia.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2011/_cr11334.pdf_