## _cr08204

## Source details

**Canonical URL:** [_cr08204](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2008/_cr08204.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2008/_cr08204.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2008/_cr08204.pdf.json)

---

### Overall resilience
- The Austrian banking system "exhibits considerable resilience against shocks."
- Resilience supported by:
  - banks that are "generally well-managed and well-capitalized";
  - a moderate attitude toward market risks and an "originate and hold" lending strategy that yields relationship banking and incentives for careful monitoring of credit risk;
  - a stable funding source in deposits, aided by a tiered structure where small banks contribute to the liquidity of the apex institution of their sector;
  - "ample liquidity" at the banking sector level.

### Main sources of risk
- Primary sources:
  - Credit risk from exposures to Central, Eastern, and Southeastern Europe (CESE) and the Commonwealth of Independent States (CIS).
  - Indirect credit risk from foreign currency (FX) lending and associated repayment vehicles (RPVs).
  - Domestic credit risk from lending in Austria.
- Observations:
  - Several large Austrian banks have substantial exposure to CESE and the CIS via foreign subsidiaries and direct nondomestic lending; returns in CESE and the CIS are "high."
  - Foreign currency lending high in absolute terms; domestic FX loan volume declining as a percentage of total loans and is "almost all" in Swiss francs rather than Japanese yen.
  - FX loans in CESE and the CIS were excluded from indirect FX stress tests due to lack of data and "remain on the increase."

### Coverage and scope of the stress tests
- Focus: the six largest Austrian banking groups: Erste, Bank Austria Creditanstalt (BA-Ca), Raiffeisen Zentralbank Oesterreich AG (RZB), Österreichische Volksbank AG (OeVAG), BAWAG P.S.K., Hypo Alpe-Adria-Bank International AG (HAA).
- Market share and scale:
  - Six groups together: "68 percent" market share in total assets in Austria (consolidated mid-2007).
  - Total bank assets in Austria: "equivalent to about 300 percent of GDP in 2006."
  - "871 credit institutions active in Austria at end-2006."
  - Market shares in CESE and CIS host markets vary "between 1 and 53 percent."
- Tests covered trading and banking books and all on– and off–balance sheet items at group level including CESE and CIS subsidiaries.
- Approaches:
  - Bottom-Up (BU): decentralized, banks using internal models; coordinated by OeNB.
  - Top-Down (TD): centralized, OeNB using supervisory data.
- All macroeconomic and single-factor market risk stress tests based on "end-June 2007" data; banks reported results in millions of euros.

### Stress test scenarios and principal outcomes
- Two macro scenarios:
  - CESE scenario: severe CESE-originated shock.
  - Global Downturn scenario: global downturn causing prolonged domestic recession.
- High-level results:
  - Both scenarios generate "substantial credit losses," causing "very low or in some cases even negative return on equity," but "capital buffers generally remained intact."
  - "All of the large banks stayed well above the 8 percent minimum capital requirements under stress."
  - Top-down analyses indicate many smaller banks (representing a very small proportion of assets) would be strained by severe domestic developments, but sectoral guarantees imply "no significant contagious defaults are predicted."
  - BU tests: high profits allow most banks to absorb most impact; only a few cases affect capital materially.

### Key numeric scenario outcomes (selected exact figures preserved verbatim)
- Sector and profit deviations under CESE shock (2007:Q3 to 2010:Q2, percentage deviations from baseline):
  - GDP Austria: -0.4, -0.4, -0.6, -0.8, -0.9, -1.0, -1.1, -1.1, -1.1, -1.1, -1.1, -1.0
  - GDP NMS-07: -5.9, -7.3, -8.1, -8.5, -8.7, -7.4, -6.1, -4.9, -3.8, -2.8, -2.0, -1.3
  - Profit before credit losses: -0.9, -2.1, -4.6, -7.2, -9.8, -12.7, -14.8, -16.4, -16.7, -16.7, -16.3, -15.9
- Austria under Global Downturn (2007:Q3 to 2010:Q2, percentage deviations from baseline):
  - GDP Austria: -0.6, -1.5, -2.4, -3.2, -4.1, -4.7, -5.2, -5.8, -6.1, -6.2, -6.2, -6.1
  - Profits before credit losses: -2.1, -4.8, -7.9, -11.3, -12.7, -14.3, -15.7, -16.9, -17.1, -17.1, -17.1, -16.9
  - Short-term interest rate (bps): 50, 146, 180, 200, 200, 200, 200, 200, 200, 200, 200, 200

### Credit-risk calibration and PD / LLP effects
- PD increases (percentage deviation from baseline):
  - CESE Scenario — PDs Austria: 11.2, 18.0, 27.6 (Year 1, Year 2, Year 3)
  - Global Downturn Scenario — PDs Austria: 15.3, 36.7, 71.3
- LLP developments in CESE/CIS (selected exact values):
  - CESE Scenario — LLPR NMS-04: 101.7, 128.9, 81.6
  - CESE Scenario — LLPR NMS-07: 134.6, 145.2, -29.1
  - CESE Scenario — LLPR SEE: 78.1, 116.4, 33.1
  - CESE Scenario — LLPR CIS: 32.0, 68.4, 114.2
- PD estimation note: annual domestic default rates increase up to 28 percent higher for CESE scenario and up to 71 percent higher in Global Downturn scenario (relative deviations).

### Loss, capital and RoE impacts — CESE vs Global Downturn (selected exact numbers)
- CESE scenario — total additional credit losses for largest six banks over three years:
  - TD results: some €10 billion.
  - BU results: €6.3 billion.
  - Total regulatory capital: some €41.4 billion.
  - Quarterly profits at mid-2007: €1.6 billion.
  - Quarterly peak average credit losses: TD €189 million (Q3 2008), BU €137 million (Q3 2008).
  - RoE (average TD): declined to 4 percent in second year; mid-2007 average RoE was 22 percent.
  - TD profit impact in capital terms: about 18 months' profit lost (one year in BU), amounting to some 2.8 percentage points of regulatory (tier I + tier II) capital assuming zero profits.
- CESE alternative severe sensitivity:
  - Total credit losses for large six banks = €11.1 billion over three years.
  - Two major banks lose approximately 1 percentage point of regulatory capital.
  - All six banks remain above the 8 percent minimum capital requirement.
  - About 4 percent of all Austrian banks (representing 1.2 percent of assets) fall below 8 percent capital; some 1 percent of banks (representing 0.2 percent of assets) fall below 4 percent capital.
- Global Downturn — domestic portfolios only — total additional credit losses for largest six banks over three years:
  - TD results: some €4.9 billion.
  - BU results: €1.6 billion.
  - Expected losses for most banks rose by between 125 and 150 percent.
  - Baseline expected losses: average €305 million.
  - RoE effects (average TD): Year 2: 13 percent (TD) versus 21 percent (BU); Year 3: 9 percent (TD) versus 19 percent (BU).
  - Profit impact in capital terms: TD suggests loss of about three quarters of a year’s profit (one quarter BU), amounting to some 1.3 percentage points of regulatory (tier I + tier II) capital.

### System-wide top-down contagion and capital adequacy findings
- Top-down system-wide CESE baseline:
  - Approximately 1.1 percent of banks would fall below the 8 percent capital ratio in year 3.
  - An additional 0.8 percent of banks would see regulatory capital fall below 4 percent.
  - Banks falling below the 8 percent limit represent 0.3 percent of total banking assets.
- Broader top-down analysis (later summary):
  - Approximately 4.6 percent of banks would fall below the 8 percent capital ratio in year 3.
  - An additional 0.6 percent of banks would see their regulatory capital fall below 4 percent.
  - In terms of assets: 1.4 percent of banks fall below the 8 percent capital requirement; an additional 0.1 percent of assets fall below the 4 percent capital level.
  - "Meanwhile, banks representing 91 percent of assets maintain a capital adequacy ratio above 10 percent."
- More severe alternative:
  - After contagious defaults, TD estimated credit losses in the large six banks amount to €5.7 billion over three years (around €0.8 billion more than baseline).
  - One major bank loses approximately 0.6 percentage point of regulatory capital; all six remain above the 8 percent minimum.
  - Around 6 percent of all Austrian banks, representing 2 percent of banking assets, would fall below the 8 percent regulatory capital requirement.
  - Some 0.7 percent of banks, representing 0.4 percent of assets, would experience capital decline to below 4 percent.

### Indirect credit risk from FX movements — exposures and calibrated shocks
- FX lending magnitudes and composition:
  - Foreign currency lending in Austria: 17.3 percent of total loans (some €48.5 billion).
  - Around 29 percent of total loans to private households are in foreign currency.
  - Currency composition: CHF dominant; JPY 3 percent; USD approximately 6 percent.
  - Over 70 percent of households FX loans have RPVs associated with them; RPVs largely life insurance policies, many with guaranteed minimum return.
- Shock assumptions used:
  - Euro versus Swiss franc: -10 percent.
  - Euro versus Japanese yen: -20 percent.
  - RPVs underperformance: 15 percent.
- Indirect FX stress-test results (system and large-bank impacts):
  - System-level impact: change in LLPs some 300 percent, implying an impact of 1.4 percent of capital of the banking system.
  - Roughly half of this impact due to FX movements; half due to underperformance of the RPV.
  - For large banks, impact ranged from 0.1 percent of capital to 1.4 percent, asset-weighted average 0.7 percent of capital.
  - EUR/JPY movements minor due to low JPY loan volume.
- Caveats:
  - FX loans in CESE and CIS excluded from these tests due to lack of data.
  - Total impact assumed to occur within a year despite many FX mortgages’ long durations.
  - Many RPVs are life insurance products with guaranteed minimum returns.
  - Currency composition shifted from JPY to CHF, reducing historical volatility vis-à-vis the euro.

### Liquidity assessment and stress-testing results
- Liquidity posture:
  - Austrian banking system "exhibits ample liquidity."
  - Many banks have stable deposit funding; tiered structure supports apex institutions.
  - Large banks have liquidity management systems of varying sophistication.
- Liquidity scenario parameters:
  - Spread between secured and unsecured euro money market rates increases by 80 bps.
  - Pool of collateral accepted in secured market (excluding government bonds) shrinks by 30 percent.
  - 30 percent of currently eligible assets fall below ECB single-A threshold and become ineligible.
  - Duration: 3 months; effects reported after 30, 60, and 90 days.
- Single-factor liquidity shocks:
  1. Decrease in market value of liquid bonds by 25 percent.
  2. Decrease in market value of equity portfolio by 35 percent.
  3. Withdrawal of 40 percent of all short-term funding.
  4. Withdrawal of 50 percent of short-term deposits of nonbank customers.
- Combined scenario specifics:
  - Bonds and equities market values fall by 20 percent and 30 percent respectively.
  - Bank-specific idiosyncratic deposit withdrawals: sight 10 percent, one-month 20 percent, three-month 30 percent.
  - Interbank funding reduction: -20 percent (overnight), -30 percent (one-month), -40 percent (three-month).
- Top-down liquidity ratio results (selected exact entries, in percent):
  - Unstressed system means:
    - Liquidity Ratio 1 Mean: 45; Standard Deviation: 18
    - Liquidity Ratio 2 Mean: 48; Standard Deviation: 23
    - Liquidity Ratio 3 Mean: 76; Standard Deviation: 32
  - Withdrawal of 40% of ST funding:
    - Liquidity Ratio 1 Mean: 30; Standard Deviation: 24
    - Liquidity Ratio 2 Mean: 36; Standard Deviation: 28
    - Liquidity Ratio 3 Mean: 73; Standard Deviation: 35
  - Withdrawal of 50% of deposits:
    - Liquidity Ratio 1 Mean: 31; Standard Deviation: 20
    - Liquidity Ratio 2 Mean: 34; Standard Deviation: 32
    - Liquidity Ratio 3 Mean: 68; Standard Deviation: 45
  - Combined scenario:
    - Liquidity Ratio 1 Mean: 19; Standard Deviation: 21
    - Liquidity Ratio 2 Mean: 15; Standard Deviation: 28
    - Liquidity Ratio 3 Mean: 58; Standard Deviation: 39
- Interpretation:
  - Single-factor shocks reduced short-term liquidity ratio 1 by some 30–35 percent when 40 percent of short-term funding or 50 percent of nonbank deposits withdrawn.
  - Scenario shock: short-term liquidity ratio decreased by some 60 percent, or even 70 percent when overnight funding included; three-month liquidity ratio decreased by 24 percent.
  - Top-down liquidity effects: no large bank would experience a liquidity shortage under shocks.

### Market-risk shocks (calibrated) and summary impacts
- Calibrated market shocks:
  - Interest Rates: parallel ±200 bps, steepening via 200 bps increase of 10-year rate.
  - Equity Prices: domestic and nondomestic -35 percent.
  - Exchange Rates: EUR depreciation/appreciation ±15 percent.
  - Implied Volatility: +200 bps, -100 bps.
- Market-risk impact (selected exact entries: change in capital adequacy ratio in percentage point):
  - Interest Rates — Parallel upward shift of euro yield curve by 200 bps:
    - BU average impact (weighted by assets): -0.16; BU dispersion: 0.20
    - TD average impact: -0.34; TD dispersion: 0.31
  - Interest Rates — Parallel downward shift of euro yield curve by 200 bps:
    - BU average impact: 0.13; BU dispersion: 0.20
    - TD average impact: 0.39; TD dispersion: 0.36
  - Equity Prices — Decrease in domestic equity prices by 35%:
    - BU average impact: -0.04; BU dispersion: 0.05
    - TD average impact: -0.09; TD dispersion: 0.05
  - Exchange Rates — Depreciation of euro by 15%:
    - BU average impact: -0.14; BU dispersion: 0.33
    - TD average impact: 0.08; TD dispersion: 0.19
- Observations:
  - Interest rate risk dominates market risks but is limited: an instantaneous 200 bps increase across the euro yield curve leads to losses corresponding to a decrease in capital of 0.2 percentage points on average against an average capital ratio of 11.5 percent before the shock.
  - Other market risks and derivative exposures are minor in aggregate; differences between BU and TD arise from aggregate reporting and delta-weighting of derivatives.

### Modeling capacity, model risk and methodological caveats
- Modeling capacities among large banks "vary," reflecting strategic focus and IRB preparation stages; banks focused on large corporates use more sophisticated credit models.
- Model risk concerns:
  - Favorable recent macroeconomic history in Austria, CESE, and the CIS may lead credit risk indicators and loan loss provisions to "underestimate risks."
  - TD approach included "many worst-case assumptions" relying on judgment that "might turn out to be too optimistic."
- Methodological caveats:
  - Lack of fully modeled interaction/feedback among institutions; no modeling of dynamic bank reactions (e.g., capital raising, "fire sales"); portfolios assumed fixed over three-year scenarios.
  - Dispersion of credit-risk models and heterogeneity of portfolios reduce cross-institutional consistency.
  - Data limitations in TD tests required assumptions on LGD, subsidiary incorporation, and corporate ratings; conservative assumptions employed.

### Key recommendations and policy measures
- Encourage and monitor development of modeling capacity in large banks:
  - More regular contacts between authorities and banks on stress testing, e.g., coordinated stress tests on a regular basis.
  - Encourage banks to develop capacity to model the link between macroeconomic environment and their credit risk and profits without relying on OeNB estimates.
  - Authorities should pay specific attention to banks’ treatment of model risk, given high growth and low perceived credit risk in CESE, the CIS, and domestically.
  - Some medium-sized banks could be encouraged to develop limited stress testing capabilities.
- Use OeNB’s off-site modeling capacity as basis for ongoing interaction with large commercial banks:
  - Support continued emphasis on model development by OeNB as a key enabling skill for dialogue on credit risk modeling and macro–credit risk links.
- Remain vigilant on indirect credit risk from exchange rate movements:
  - Authorities have focused on these risks through publications, public information campaigns, and minimum standards for granting and managing FX loans and loans with RPVs.
  - Continued monitoring warranted given still-high domestic FX loans, large volume of FX loans abroad held by Austrian subsidiaries, and stress-test results showing substantial impact from exchange rate movements.
- Remain alert to concentrated exposure to CESE and CIS and intensify cross-border cooperation:
  - Supervisors should continue to discuss these issues with banks and further intensify cross-border supervisory cooperation.
- Additional model-development actions:
  - Investigate modeling PDs as a function of macroeconomic data, possibly via structural models where interest rates play a prominent role.
  - Explore estimating absolute increases in PDs in addition to relative increases.
  - Perform reconciliation exercises between off-site TD estimates and banks’ BU estimates to increase confidence in models.

*Source: Executive Summary and sections as contained in _cr08204.*

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

### Executive Summary

### Overall resilience
- The Austrian banking system "exhibits considerable resilience against shocks."
- Resilience supported by:
  - banks that are "generally well-managed and well-capitalized";
  - a moderate attitude toward market risks and an "originate and hold" lending strategy that yields relationship banking and incentives for careful monitoring of credit risk;
  - a stable funding source in deposits, aided by a tiered structure where small banks contribute to the liquidity of the apex institution of their sector;
  - "ample liquidity" at the banking sector level.

### Main sources of risk
- Primary sources of risk identified:
  - Credit risk from exposures to Central, Eastern, and Southeastern Europe (CESE) and the Commonwealth of Independent States (CIS).
  - Indirect credit risk from foreign currency (FX) lending and associated repayment vehicles (RPVs).
  - Domestic credit risk from lending in Austria.
- Several large Austrian banks have substantial exposure to CESE and the CIS via foreign subsidiaries and direct nondomestic lending. Returns in CESE and the CIS are noted as "high."
- Foreign currency lending remains high in absolute terms:
  - Domestic FX loan volume is declining as a percentage of total loans and is "almost all" in Swiss francs rather than Japanese yen.
  - FX loans in CESE and the CIS were not included in the indirect FX stress tests due to lack of data and "remain on the increase."

### Coverage and scope
- Stress tests centered on six largest Austrian banking groups: Erste, Bank Austria Creditanstalt (BA-Ca), Raiffeisen Zentralbank Oesterreich AG (RZB), Österreichische Volksbank AG (OeVAG), BAWAG P.S.K., and Hypo Alpe-Adria-Bank International AG (HAA).
- The six banking groups together have a "68 percent" market share in terms of total assets in Austria (consolidated data from mid-2007).
- Total bank assets in Austria were "equivalent to about 300 percent of GDP in 2006," with "871 credit institutions active in Austria at end-2006."
- Market shares in CESE and CIS host markets vary "between 1 and 53 percent."
- Tests covered all major portfolios, including trading and banking books and all on– and off–balance sheet items, performed on a group level including CESE and CIS subsidiaries.
- Two approaches used:
  - Bottom-Up (BU): decentralized, performed by largest banks using internal models and coordinated by OeNB.
  - Top-Down (TD): centralized, performed by OeNB using supervisory data.
- All macroeconomic and single-factor market risk stress tests are based on "end-June 2007" data.

### Stress test scenarios and outcomes
- Two macroeconomic scenarios:
  - A severe CESE-originated shock (CESE scenario).
  - A global downturn causing a prolonged domestic recession (Global Downturn scenario).
- Results:
  - Both scenarios generate "substantial credit losses," causing "very low or in some cases even negative return on equity," but "capital buffers generally remained intact."
  - "All of the large banks stayed well above the 8 percent minimum capital requirements under stress."
  - Top-down analyses indicate "a substantial number of smaller banks (albeit representing only a very small proportion of banking system assets) would be strained" by severely adverse domestic macroeconomic developments; many of these banks "enjoy a form of guarantee from the sector" so actual default would likely not occur, and "no significant contagious defaults are predicted."
  - The BU tests show high profits would allow most banks to absorb most of the impact; only in a few cases would losses affect capital.

### Indirect credit risk from FX movements
- Stress tests for indirect credit risk combined negative shocks to exchange rates and to the performance of repayment vehicles connected with FX loans.
- Findings:
  - System-level resilience confirmed, but "the impact on the system as a whole and on some of the large banks is considerable."
  - The impact reflects primarily "the large outstanding volume of FX loans."
  - Domestic FX loan volume is declining as a percentage of total loans and is "almost all" in Swiss francs rather than Japanese yen; nevertheless, stress tests show a substantial impact, in part due to "conservative modeling assumptions."
  - Several small banks (representing a very small part of assets) would experience severe strain but mostly enjoy sectoral guarantees; therefore "no systemic impact would be expected."
  - The indirect FX stress tests were carried out off site by the authorities and "have not been compared with estimates from the banks themselves."
  - FX loans in CESE and the CIS were excluded from these tests due to lack of data.

### Liquidity
- The Austrian banking system "exhibits ample liquidity."
- Many banks have stable deposit funding; tiered banking structure supports liquidity of apex institutions.
- Large banks have liquidity management systems of varying sophistication.
- Liquidity stress tests and observed robustness during recent credit market turmoil "suggest that banks follow a prudent approach toward liquidity."
- Banks expect liquidity management to gain prominence given the prolonged market turmoil, but none of the large banks foresee "major strains over and above the generally higher market price for liquidity." This confidence is judged "justified, at least for the system as a whole."

### Modeling capacity and model risk
- Modeling capacities among larger banks "vary," reflecting strategic focus and different stages of preparation for IRB approval.
  - Banks focusing on large corporates generally used more sophisticated credit risk models.
- Banks reported that most risks were covered and that the scenarios were "severe but plausible," though one major bank noted internal sensitivity tests can show a more severe impact on credit risk indicators.
- Given favorable macroeconomic developments in Austria, CESE, and the CIS over recent years, credit risk indicators and loan loss provisions based on that period may "underestimate risks," implying model risk.
- The TD approach included "many worst-case assumptions" that "necessarily rely to a large degree on judgment," which "might turn out to be too optimistic."

### Recommendations
- Encourage and monitor development of modeling capacity in large banks:
  - More regular contacts between authorities and banks on stress testing, e.g., coordinated stress tests on a regular basis.
  - Encourage banks to develop capacity to model the link between macroeconomic environment and their credit risk and profits without relying on OeNB estimates.
  - Authorities should pay specific attention to banks’ treatment of model risk, given high growth and low perceived credit risk in CESE, the CIS, and domestically.
  - Some medium-sized banks could be encouraged to develop limited stress testing capabilities.
- Use OeNB’s sophisticated off-site modeling capacity as basis for ongoing interaction with large commercial banks on stress testing:
  - Support continued emphasis on model development by OeNB as a key enabling skill for dialogue on credit risk modeling and macro–credit risk links.
- Remain vigilant on indirect credit risk from exchange rate movements:
  - Authorities have focused on these risks through publications, public information campaigns, and minimum standards for granting and managing FX loans and loans with RPVs.
  - Continued monitoring is warranted given the still-high level of domestic FX loans, large volume of FX loans abroad held by Austrian-owned subsidiaries, and stress test results showing substantial impact from exchange rate movements.
- Remain alert to concentrated exposure to CESE and CIS and intensify cross-border cooperation:
  - While banks are generally well-diversified within the region, a general or contagious downturn would significantly affect them and might lead to funding problems even though most funding is through deposits.
  - Supervisors should continue to discuss these issues with banks and further intensify cross-border supervisory cooperation.

*Source: Executive Summary, _cr08204 - Executive Summary.*

### 6.      The stress testing exercise aims to include all major risks from macroeconomic

### _cr08204 - 6.      The stress testing exercise aims to include all major risks from macroeconomic

### Overview of the stress testing exercise
- Two multi-factor macroeconomic scenario stress tests:
  - Scenario 1: large capital account adjustment and concomitant recession in CESE.
  - Scenario 2: sharp global downturn resulting in a prolonged domestic recession.
- Sensitivity analyses performed around both base stress scenarios.
- Global downturn scenario included contagion analysis through the interbank market.
- Single-factor market-risk tests (interest rates, equity prices, exchange rates, implied volatility of options) performed by all institutions.
- Short-term vulnerabilities assessed focusing on liquidity (qualitative and limited quantitative analysis); operational risk was not assessed.
- Data basis: end-June 2007; banks reported results in millions of euros.

### Methodology — Bottom-Up (BU) approach
- Credit and market risks tested using institutions’ own internal risk models.
- BU tests used end-June 2007 data and projections out to 2010; banks reported in millions of euros.
- To enhance consistency across heterogeneous credit models, OeNB provided:
  - Estimates for relative changes in probabilities of default (PDs) and loan loss provisions (LLPs) under the macroeconomic scenarios.
  - A profile for profits before credit losses under the macroeconomic scenarios (estimated using bank profit data from the Asia crisis).
- Short-term liquidity assessment:
  - Focused on the six large banks.
  - Consisted of a questionnaire and a scenario shocking asset liquidity over 30, 60, and 90 days.

### Methodology — Top-Down (TD) approach
- TD stress tests based solely on OeNB modeling of supervisory data; performed on end-June 2007 data; results in millions of euros.
- Allowed analysis of the entire Austrian banking system using supervisory data.
- Contagion analysis for the global downturn used an integrated model of the Austrian interbank market combining standard risk techniques with a network model (covering all maturities) based on supervisory filings and OeNB’s Systemic Risk Monitor (SRM).

### Methodological caveats (general and specific)
- General caveats:
  - Lack of interaction/feedback modeling between financial institutions limits full externality assessment.
  - Partial remedies: sensitivity analyses, liquidity scenario, interbank contagion analysis via network model.
- Scenario-specific caveats:
  - Starting position of the banking system: low defaults and high profit buffers — unusually strong.
  - Results would worsen under more severe scenarios (e.g., sharper current account adjustment, large CESE currency swings).
  - Potential contagion among CESE and CIS could produce larger spillovers to Austria than historically observed.
  - Feedback effects of scenarios on economic activity and asset prices in CESE and CIS were not assessed.
  - Severe CESE downturn might cause funding problems for Austrian banks in the interbank market.
- Reaction modeling:
  - No modeling of banks’ dynamic reactions (e.g., capital raising, “fire sales”); exercise assumes fixed portfolios over the three-year scenarios, though banks could report alternate profit forecasts.
- Specific caveats:
  - Dispersion of credit-risk models and heterogeneity of portfolios may reduce realism of impacts.
  - Uniform assumed profit development before credit-risk losses is an oversimplification given geographic diversification.
  - Data limitations in TD tests required assumptions on LGD, subsidiary incorporation, and corporate ratings; conservative assumptions employed.

### Shocks and short-term vulnerabilities — Macroeconomic scenarios
- Both scenarios cover a three-year horizon starting in 2007:Q3, with models generating baseline quarterly variables for 2007:Q3 to 2010:Q2 and for the stressed paths.
- CESE scenario features a confidence crisis in CESE with:
  - Sharp decrease of about half of the current account deficit in the countries involved over one year.
  - Disabled monetary policy reaction (CESE countries continue to shadow the euro).
  - Real effects: level of GDP up to 9 percent lower in 2008 in the two New Member States that joined the EU in 2007.
  - Spillovers to Austria:
    - Exports decline up to 1.5 percent.
    - Austrian GDP around 1 percent lower than baseline (excluding indirect effect via bank exposures).
    - Austrian bank profits before credit losses assumed to decline by up to 17 percent from current levels.
- Global downturn scenario assumptions:
  - Sharp downturn in economies of Austria’s main trading partners in Western Europe.
  - Confidence effects: gradual rise in risk premia of 200 bps.
  - Gradual rise in domestic household savings rate of 1.5 percentage points.
  - Domestic consequences (annual GDP growth): 2.8, -0.4, and -0.1 percent for 2007–2009 respectively.
  - Bank profits assumed to decline up to 17 percent from current levels.
  - Scenario focuses on domestic effects and does not incorporate indirect effects on CESE.

### Key scenario tables (selected numeric outcomes preserved verbatim)
- Table 1 — Austria: Real GDP and Profit Development Under the CESE Shock (Percentage deviations from baseline levels)
  - GDP Austria: -0.4, -0.4, -0.6, -0.8, -0.9, -1.0, -1.1, -1.1, -1.1, -1.1, -1.1, -1.0 (2007 Q3 to 2010 Q2)
  - GDP NMS-04: -1.7, -2.7, -3.5, -4.0, -4.4, -4.2, -3.9, -3.5, -3.2, -2.9, -2.7, -2.5
  - GDP NMS-07: -5.9, -7.3, -8.1, -8.5, -8.7, -7.4, -6.1, -4.9, -3.8, -2.8, -2.0, -1.3
  - GDP SEE: -2.1, -3.2, -3.9, -4.4, -4.8, -4.5, -4.1, -3.7, -3.3, -2.9, -2.6, -2.4
  - GDP CIS: -0.1, 0.0, -0.1, -0.2, -0.3, -0.4, -0.4, -0.5, -0.6, -0.6, -0.6, -0.6
  - Profit before credit losses: -0.9, -2.1, -4.6, -7.2, -9.8, -12.7, -14.8, -16.4, -16.7, -16.7, -16.3, -15.9
- Table 2 — Austria: Real GDP, Profit, and Interest Rate Developments Under the Global Downturn Shock (Percentage deviations from baseline levels)
  - GDP Austria: -0.6, -1.5, -2.4, -3.2, -4.1, -4.7, -5.2, -5.8, -6.1, -6.2, -6.2, -6.1 (2007 Q3 to 2010 Q2)
  - Profits before credit losses: -2.1, -4.8, -7.9, -11.3, -12.7, -14.3, -15.7, -16.9, -17.1, -17.1, -17.1, -16.9
  - Short-term interest rate (bps): 50, 146, 180, 200, 200, 200, 200, 200, 200, 200, 200, 200

### Credit-risk calibration and results
- OeNB provided estimated relative changes in domestic PDs and LLPs consistent with scenarios.
- PD estimation:
  - Based on regression of default rates of Austrian corporates on GDP development, unemployment rate, fixed capital investment, oil price, and some role for interest rates.
  - Annual domestic default rates increase up to 28 percent higher for the CESE scenario and up to 71 percent higher in the global downturn scenario (relative deviations).
- LLP developments for CESE and CIS modeled as function of GDP due to data restrictions:
  - LLPs up to 145 percent higher than baseline for Romania and Bulgaria.
  - Around 130 percent increase for the NMS 2004.
  - 116 percent increase for Southeastern European countries.
- Table 3 — Austria: Credit Risk Indicators (Percentage deviation from baseline)
  - CESE Scenario — PDs Austria: 11.2, 18.0, 27.6 (Year 1, Year 2, Year 3)
  - CESE Scenario — LLPR NMS-04: 101.7, 128.9, 81.6
  - CESE Scenario — LLPR NMS-07: 134.6, 145.2, -29.1
  - CESE Scenario — LLPR SEE: 78.1, 116.4, 33.1
  - CESE Scenario — LLPR CIS: 32.0, 68.4, 114.2
  - Global Downturn Scenario — PDs Austria: 15.3, 36.7, 71.3

### Loss reporting and top-down sensitivity/contagion analyses
- Banks reported credit losses in millions of euro; domestic expected losses computed under stressed PDs; CESE expected losses based on increases in LLPs.
- Some banks also reported unexpected credit losses.
- Results related to profits (at assumed reduced levels) and capital (regulatory own funds).
- Top-down models used to:
  - Calculate sensitivity around stress scenarios.
  - Analyze contagion using adapted SRM integrating macroeconomic credit-risk model with interbank network model.
- CESE scenario sensitivity: alternative runs applied credit-risk deterioration patterns of Romania and Bulgaria to some/all CESE and CIS countries combined with a domestic household savings rate increase of 2 percentage points.
- Global downturn sensitivity: additional Austrian household savings rate increase of 2 percentage point (i.e., total increase of 3.5 percentage point compared to baseline).

### Market-risk shocks (summary of calibrated shocks)
- Interest Rates:
  - Parallel upward shift of euro yield curve by 200 bps
  - Parallel downward shift of euro yield curve by 200 bps
  - Steepening of euro yield curve through 200 bps increase of 10-year rate
- Equity Prices:
  - Decrease in domestic equity prices by 35 percent
  - Decrease in nondomestic equity prices by 35 percent
- Exchange Rates:
  - Depreciation of euro by 15 percent
  - Appreciation of euro by 15 percent
- Implied Volatility:
  - Increase in implied volatility by 200 bps
  - Decrease in implied volatility by 100 bps
- Credit spread risk was not assessed due to very limited exposure and lack of tools.
- TD sensitivity analyses included interest rate shifts and exchange rate shocks to the USD, JPY, CHF, and GBP; estimated impacts (losses relative to capital) were negligible and omitted from BU workload.

### Modeling notes and limitations relevant for interpretation
- Market-risk models generally consistent across institutions; credit-risk models more diverse, affecting cross-institutional consistency.
- PDs estimated using historical default frequencies from 1970 to mid 2007; example: annualized PD for the overall Austrian economy for 2007 Q2 was 2.7 percent.
- Top-down contagion analysis assumption: all banks falling below a 4 percent capital adequacy ratio are assumed to default on their interbank liabilities; the 4 percent capital interpreted as costs associated with a default.

*Source: OeNB (as contained in _cr08204 - 6. The stress testing exercise aims to include all major risks from macroeconomic).*

### 28.      In addition to the standard market risk, a top-down analysis of indirect credit

### _cr08204 - 28.      In addition to the standard market risk, a top-down analysis of indirect credit

### Indirect credit risk from exchange rate movements (FX)
- Foreign currency lending in Austria stands at 17.3 percent of total loans (some €48.5 billion).
- Around 29 percent of total loans to private households are in foreign currency.
- Currency composition of foreign currency lending:
  - Swiss franc (CHF): much of the total foreign currency lending (dominant).
  - Japanese yen (JPY): 3 percent.
  - U.S. dollars (USD): approximately 6 percent.
- Households are mostly unhedged; some households in western Austria have income in Swiss franc.
- Much of private household foreign currency lending is for mortgages; over 70 percent of households FX loans have RPVs associated with them.
- RPVs largely take the form of life insurance policies, many with a guaranteed minimum return; a 15 percent shortfall in RPV performance is considered severe.
- Shock assumptions used in analysis:
  - Euro versus Swiss franc: -10 percent.
  - Euro versus Japanese yen: -20 percent.
  - RPVs underperformance: 15 percent.
- Results are presented in terms of percentage point of regulatory capital.

### Liquidity risk and liquidity stress testing
- Liquidity stress tests focused on the six largest banks and included both bottom-up (BU) and top-down (TD) components.
- Banks were surveyed on compliance with BIS Sound Practices for Managing Liquidity in Banking Organizations (BIS, 2000) and on regular liquidity stress testing practices.
- Specific liquidity scenario parameters:
  - Spread between secured and unsecured euro money market rates increases by 80 bps.
  - Pool of collateral accepted in the secured market (excluding government bonds) shrinks by 30 percent.
  - 30 percent of currently eligible assets (excluding government bonds) fall below the ECB single-A quality threshold and become ineligible as collateral for central bank loans.
  - Duration: 3 months; effects reported after 30, 60, and 90 days.
- Liquidity ratios analyzed (TD):
  - Liquidity ratio 1: liquid assets to short-term liabilities (short-term liabilities = residual maturity up to three months; liquid assets = cash reserves + debt instruments accepted for ECB refinancing + listed equities and bonds).
  - Liquidity ratios 2 and 3: broader ratios adding respectively overnight or short-term (up to three months remaining maturity) interbank and nonbank assets to liquid assets; short-term nonbank assets are weighted by 0.5.
- Single-factor liquidity shocks analyzed:
  1. Decrease in market value of liquid bonds by 25 percent.
  2. Decrease in market value of equity portfolio by 35 percent.
  3. Withdrawal of 40 percent of all short-term funding.
  4. Withdrawal of 50 percent of short-term deposits of nonbank customers.
- Combined scenario:
  - Decrease in market value of bonds and equities by 20 percent and 30 percent, respectively.
  - Bank-specific idiosyncratic shock per bank: nonbank customers withdraw 10 percent of sight deposits, 20 percent of one-month deposits, and 30 percent of three-month deposits.
  - Interbank funding reduction assumed to be -20 percent (overnight), -30 percent (one-month), and -40 percent (three-month).

### Results — Overview
- Banks withstood macroeconomic shocks overall, but some weaknesses emerged.
- Credit losses under scenarios were substantial; assumed baseline profitability before credit losses could decline by up to 17 percent relative to baseline.
- Given high baseline profitability, banks generally absorbed losses through profits; RoEs declined sharply but effects on regulatory capital were minor.
- Variation across banks: effects on the group of six large banks varied substantially; a limited number of small banks outside the large six fell below the 8 percent regulatory capital standard, but this did not result in defaults via interbank contagion.
- Market risks were generally modest; Austrian banks hold only small active positions.
- Liquidity stress tests indicated large banks would not see major strains beyond generally higher market prices for liquidity.

### The CESE scenario — losses, concentration, and capital effects
- Total additional credit losses for the largest six banks over a three year horizon:
  - TD results: some €10 billion.
  - BU results: €6.3 billion.
- Comparison metrics:
  - Total regulatory capital: some €41.4 billion.
  - Quarterly profits at mid-2007: €1.6 billion.
- Loss concentration: credit risk concentrated in banks with large exposures to the CESE and CIS region.
- Quarterly peak average credit losses:
  - TD: €189 million in Q3 2008.
  - BU: €137 million in Q3 2008.
- RoE effects (average TD estimates):
  - RoE declined to 4 percent in the second year of the scenario.
  - Mid-2007 average RoE was 22 percent; large six banks exhibited RoEs between 9 and 28 percent in mid-2007.
- Profit impact expressed in regulatory capital terms:
  - TD estimates imply about 18 months’ profit lost (one year in BU estimates), amounting to some 2.8 percentage points of regulatory (tier I + tier II) capital assuming zero profits.
- Reasons TD > BU:
  1. TD assumes 100 percent LGD for uncollateralized part of credit portfolio; banks use internal LGD estimates or standard 45 percent.
  2. Banks may classify loans to the public sector under “services,” yielding lower PDs in BU; OeNB applies higher PDs based on corporate historic default frequencies.
  3. BU includes full consolidation of majority-owned subsidiaries; TD assumes losses proportional to ownership percentage.
  4. TD uses consistent reported data across banks; BU portfolios differ by bank.
  5. TD PD estimates may cover longer, more conservative historical periods; BU PDs may be optimistic given favorable recent macroeconomic history.
- Sensitivity / alternative CESE scenarios:
  - Severe alternative combining CESE/CIS shock aligned with Romania and Bulgaria plus household savings rate in Austria shocked upward by 2 percentage points.
  - Under this alternative, total credit losses for large six banks = €11.1 billion over three years (around €1 billion more than baseline CESE scenario).
  - In this alternative:
    - Two major banks lose approximately 1 percentage point of regulatory capital.
    - All six banks remain above the 8 percent minimum capital requirement.
    - About 4 percent of all Austrian banks (representing 1.2 percent of banking assets) would fall below the 8 percent regulatory capital requirement.
    - Some 1 percent of banks (representing 0.2 percent of assets) would experience capital decline to below 4 percent.
  - Most troubled banks would receive support within their sub-sector, preventing actual defaults.
- Top-down system-wide results under CESE baseline:
  - Approximately 1.1 percent of banks would fall below the 8 percent capital ratio in year 3.
  - An additional 0.8 percent of banks would see regulatory capital fall below 4 percent.
  - Banks falling below the 8 percent limit represent 0.3 percent of total banking assets.

### The Global Downturn scenario — domestic impact
- Total additional credit losses for the largest six banks over a three year horizon (domestic portfolios only):
  - TD results: some €4.9 billion.
  - BU results: €1.6 billion.
- Comparison metrics:
  - Total regulatory capital: some €41.4 billion.
  - Quarterly profits at mid-2007: €1.6 billion.
- Losses are smaller than CESE scenario because only Austrian portfolios were shocked (excluding direct cross-border lending).
- Expected losses for most banks rose by between 125 and 150 percent.
- Baseline level of expected losses is low: average €305 million; thus total impact was moderate.
- Time pattern: higher quarterly losses as the scenario progresses, reflecting a prolonged downturn with no upturn toward end of horizon.
- RoE effects (average TD estimates):
  - Year 2: 13 percent (TD) versus 21 percent (BU).
  - Year 3: 9 percent (TD) versus 19 percent (BU).
  - Half of large six banks: RoE below 10 percent in year 2.
  - Year 3: four out of six banks have RoE below 5 percent; one bank exhibits a loss.
- Profit impact expressed in regulatory capital terms:
  - TD estimates suggest loss of about three quarters of a year’s profit (one quarter BU), amounting to some 1.3 percentage points of regulatory (tier I + tier II) capital.
- TD versus BU divergence and drivers:
  - TD yields larger impact than BU for this scenario.
  - Additional TD assumption: for customers with loans at more than one bank, the TD model applies the lowest (most risky) rating across banks; banks use internal ratings that may be less conservative.
  - Inclusion in TD of already defaulted loans (reflected in reported provisions) vs. some BU treatments that excluded such loans affects results.
  - Differences in LGD assumptions and portfolio definitions further explain TD > BU differences.
  - Replacing conservative TD assumptions with BU-like assumptions (but keeping higher LGDs) reduces TD total losses to €2.8 billion versus €1.6 billion in BU.

*Sources: OeNB and banks' calculations; contents as provided in the source text.*

### 46.      A top-down analysis for the entire Austrian banking system reveals that a

### _cr08204 - 46.      A top-down analysis for the entire Austrian banking system reveals that a

### Systemic credit-impact findings (top-down analysis)
- Approximately 4.6 percent of the banks would fall below the 8 percent capital ratio in year 3.
- An additional 0.6 percent of banks would see their regulatory capital fall below 4 percent.
- In terms of assets:
  - 1.4 percent of banks fall below the 8 percent capital requirement.
  - Banks representing an additional 0.1 percent of assets fall below the 4 percent capital level.
- Most small banks under strain are expected to receive support from their sub-sector, preventing actual defaults; under the conservative assumption of no support:
  - Banks falling below 4 percent capital are assumed to default on obligations (with 100 percent LGD).
  - Interbank contagion would cause default of banks representing an additional 0.3 percent of system assets.
  - Total number of banks representing 0.4 percent of assets would fall below 4 percent capital after contagion.
  - Almost 5 percent of banks (representing 1.4 percent of system assets) would fall below the 8 percent capital requirement.
- Note: "Meanwhile, banks representing 91 percent of assets maintain a capital adequacy ratio above 10 percent."

### Alternative (more severe) downturn scenario
- Scenario assumption: a 2 percentage point higher Austrian household savings rate on top of the 1.5 percent higher household savings rate in the baseline stress scenario.
- After contagious defaults:
  - TD estimated credit losses in the large six banks amount to €5.7 billion over three years.
  - This represents around €0.8 billion in additional losses compared to the baseline stress scenario.
  - One major bank loses approximately 0.6 percentage point of regulatory capital, but all six remain above the 8 percent minimum capital requirement.
  - Around 6 percent of all Austrian banks, representing 2 percent of banking assets, would fall below the 8 percent regulatory capital requirement.
  - Some 0.7 percent of banks, representing 0.4 percent of assets, would experience a capital decline to below 4 percent.

### Market risks (summary of TD and BU results)
- Interest rate risk dominates market risks but is limited:
  - An instantaneous 200 bps increase across the entire euro yield curve leads to losses corresponding to a decrease in capital of 0.2 percentage points on average, against an average capital ratio of 11.5 percent before the shock.
- Table 7: Market Risk Scenarios (Change in capital adequacy ratio in percentage point) — selected exact entries
  - Interest Rates — Parallel upward shift of euro yield curve by 200 bps:
    - BU average impact (weighted by assets): -0.16
    - BU dispersion (unweighted standard deviation): 0.20
    - TD average impact: -0.34
    - TD dispersion: 0.31
  - Interest Rates — Parallel downward shift of euro yield curve by 200 bps:
    - BU average impact: 0.13; BU dispersion: 0.20
    - TD average impact: 0.39; TD dispersion: 0.36
  - Interest Rates — Steepening of euro yield curve through 200 bps increase of 10 year rate:
    - BU average impact: -0.08; BU dispersion: 0.10
    - TD average impact: -0.23; TD dispersion: 0.23
  - Equity Prices — Decrease in domestic equity prices by 35%:
    - BU average impact: -0.04; BU dispersion: 0.05
    - TD average impact: -0.09; TD dispersion: 0.05
  - Exchange Rates — Depreciation of euro by 15%:
    - BU average impact: -0.14; BU dispersion: 0.33
    - TD average impact: 0.08; TD dispersion: 0.19
  - Exchange Rates — Appreciation of euro by 15%:
    - BU average impact: 0.19; BU dispersion: 0.31
    - TD average impact: -0.08; TD dispersion: 0.19
  - Implied Volatility — Increase of implied volatility by 200bps:
    - BU average impact: 0.00; BU dispersion: 0.01
  - Implied Volatility — Decrease of implied volatility by 100bps:
    - BU average impact: 0.00; BU dispersion: 0.00
- Observations:
  - Other market risks, including exposure to derivatives, are minor compared to interest rate risk, except for direct exchange rate risk driven by one bank holding capital in foreign subsidiaries in local currency.
  - Differences between BU and TD analyses partly arise from aggregate reported data (OeNB) versus banks’ individual-position data; delta-weighting of derivatives in aggregates can distort results for structured products.

### Indirect credit risk induced by foreign exchange rate risk
- Stress scenario: CHF/EUR rate shocked downward by 10 percent and repayment vehicle (RPV) assumed to perform 15 percent worse than baseline.
  - Impact on LLPs was some 300 percent, implying an impact of 1.4 percent of capital of the banking system.
  - Roughly half of this impact due to FX movements; half due to underperformance of the RPV.
  - For large banks, impact ranged from 0.1 percent of capital to 1.4 percent, with an asset-weighted average impact of 0.7 percent of capital.
  - EUR/JPY exchange rate movements had minor impact due to low volume of JPY loans.
- Modeling caveats:
  - Total impact from exchange rate movements was assumed to occur within a year despite many FX mortgage loans’ long durations.
  - Many RPVs are life insurance products with guaranteed minimum returns.
  - Currency composition of domestic FX loans has shifted from JPY to CHF, reducing historical volatility vis-à-vis the euro.
- Recommendation: FX loans remain an area for continued vigilance for banks, supervisors, and financial stability authority; monitoring of indirect credit risk from FX loans in foreign subsidiaries needs improvement.

### Liquidity risk and stress-test results
- Compliance with BIS principles for Assessment of Liquidity Management in Banking Organizations is good; banks perform simple liquidity stress tests regularly.
- Banks reported liquidity positions would remain comfortable under the liquidity stress scenario due to:
  - Most interbank borrowing being secured.
  - Large stock of eligible collateral such that even a 30 percent reduction would leave adequate collateral.
- Top-down liquidity effects: no large bank would experience a liquidity shortage under shocks.
- Table 8: Liquidity Ratios Stress Test Results (In percent) — selected exact entries
  - Unstressed system means:
    - Liquidity Ratio 1 Mean: 45
    - Liquidity Ratio 2 Mean: 48
    - Liquidity Ratio 3 Mean: 76
    - Standard Deviation: 18; 23; 32 respectively
  - Withdrawal of 40% of ST funding:
    - Liquidity Ratio 1 Mean: 30; Standard Deviation: 24
    - Liquidity Ratio 2 Mean: 36; Standard Deviation: 28
    - Liquidity Ratio 3 Mean: 73; Standard Deviation: 35
  - Withdrawal of 50% of deposits:
    - Liquidity Ratio 1 Mean: 31; Standard Deviation: 20
    - Liquidity Ratio 2 Mean: 34; Standard Deviation: 32
    - Liquidity Ratio 3 Mean: 68; Standard Deviation: 45
  - Scenario (combined shocks):
    - Liquidity Ratio 1 Mean: 19; Standard Deviation: 21
    - Liquidity Ratio 2 Mean: 15; Standard Deviation: 28
    - Liquidity Ratio 3 Mean: 58; Standard Deviation: 39
- Single-factor shocks reduced short-term liquidity ratio 1 by some 30–35 percent when 40 percent of short-term funding or 50 percent of nonbank deposits is withdrawn.
- Scenario shock: short-term liquidity ratio decreased by some 60 percent, or even 70 percent when overnight funding is included; three-month liquidity ratio decreased by 24 percent.

### Qualitative assessment of risk management and modeling capacity
- Bank-level discussions revealed:
  - Variation in credit-risk modeling capacity across large banks; banks focused on large corporates use more sophisticated models.
  - Some banks would prefer macroeconomic inputs rather than PDs; OeNB provided common relative changes to PDs and LLPs to maintain consistency.
  - Banks generally found scenarios realistic and severe, though some internal tests show more severe PD impacts due to low current PD levels and use of relative (not absolute) PD changes.
  - Model risk concern: recent high-growth environment has produced low credit-risk indicators, potentially rendering model-based estimates somewhat optimistic.
- Recommendation highlights:
  - Monitor development of modeling capacity in large banks closely.
  - Encourage banks to further develop capacity to link macroeconomic environment to credit risk and profits without relying solely on OeNB-provided PD and profit estimates.
  - Use OeNB’s off-site modeling capacity as a basis for ongoing interaction and dialogue with banks on stress testing and credit-risk modeling.
  - Investigate modeling PDs as a function of macroeconomic data, possibly via structural models where interest rates play a prominent role.
  - Explore estimating absolute increases in PDs in addition to relative increases.
  - Perform reconciliation exercises between off-site TD estimates and banks’ BU estimates to increase confidence in models.

### Key recommendations (concise)
- Continue emphasis on model development at the OeNB and structured dialogue with banks on stress testing.
- Monitor and maintain vigilance on indirect credit risk from FX loans, including in foreign subsidiaries.
- Encourage banks to develop macro-linked credit-risk modeling capacity and consider absolute PD estimation approaches.
- Reconcile off-site conservative assumptions with bank-level assumptions periodically to validate stress-test outcomes.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2008/_cr08204.pdf*

### 65.      Other risks that require vigilance follow from the concentrated exposure of the

### 65.      Other risks that require vigilance follow from the concentrated exposure of the

### Concentrated exposure and supervisory implications
- Austrian banking systems have concentrated exposure to CESE and the CIS.
- While banks are generally well-diversified within this region, a general or contagious downturn in the region would significantly affect the banks.
- A severe downturn in CESE might lead to funding problems for the Austrian banks in the interbank market, even though most of their funding is through deposits.
- If banks respond by restricting credit to the affected countries, further negative macroeconomic effects might result.
- Policy implications:
  - Continued vigilance by both the banks and the supervisor is warranted.
  - Further intensification of cross-border supervisory cooperation is recommended.

### Note on structural modeling
- A structural model analogous to modeling indirect credit risk from FX movements can be considered, where higher interest rates reduce disposable incomes and move households and corporates into higher risk brackets.

---

### APPENDIX I: MODELING CREDIT-RISK MEASURES FROM MACRO FACTORS

### Purpose and approach
- To assess the effect of a macroeconomic stress scenario on credit risk, link a credit-risk measure to macroeconomic variables.
- For the Austrian credit portfolio: econometric model linking macro-variables to PDs of customers in 10 economic sectors.
- For CESE: data limitations prevent modeling PDs or NPLs; instead, LLPs (loan loss provisions) available from subsidiaries are modeled as a function of macroeconomic variables.

### A. Domestic Credit Risk — Data and segmentation
- PDs are defined as the ratio of the number of default firms to the number of firms in each sector.
- Sector definitions with number of observed firms in June 2007 (parentheses):
  - Basic: Agriculture and basic Industries (7,355)
  - Construction (26,080)
  - Energy (840)
  - Financial Services (6,092)
  - Private (885)
  - Production (22,024)
  - Services (77,162)
  - Tourism (22,263)
  - Trading (54,933)
  - Others
  - Overall (229,003)
- For sectors Energy, Private and Others (small firm counts), a model referring to the overall Austrian economy is used.

### A. Domestic Credit Risk — Model specification
- Average default probability p_{t,s} in sector s at time t is modeled as a logistic function of a sector-specific macroeconomic index y_{t,s}:
  - p_{t,s} = 1 / (1 + e^{-y_{t,s}})
- Observed macroeconomic indices y_{t,s} are calculated by the inverse logistic transform of historical PDs:
  - y_{t,s} = ln(1/p_{t,s} - 1)
- Because some y_{t,s} series are nonstationary and may have structural breaks, ordinary least squares regression explains changes in y_{t,s} by changes in macroeconomic variables.
- To adjust for seasonal effects, logarithmic changes of moving averages over four periods of macroeconomic variables are used as explanatory variables; PDs are adjusted by moving average over four periods.
- Regression model for each sector s:
  - Δy_{t,s} = β_{s,0} + β_{s,1} x_{t,1,s} + ... + β_{s,K} x_{t,K,s} + ε_{t,s}
  - where Δy_{t,s} = y_{t,s} - y_{t-1,s}, x_{t,k,s} are macro variables, and ε_{t,s} ~ N(0, σ_s^2).
- Forecasting sectoral PDs:
  - ŷ_{t,s} = ŷ_{t-1,s} + Δŷ_{t,s}; p̂_{t,s} = 1 / (1 + e^{-ŷ_{t,s}})

### Model selection and performance
- Potential macro-factors divided into seven groups; for each multivariate regression only one macro-variable from each group is integrated.
- Initial set of 27 macroeconomic variables considered (listed below in Macro variables section).
- For five sectors (basic industries, energy, financial services, private households, and others) no reasonable sectoral model could be found; the aggregated PD of the Austrian economy was applied for these sectors.
- The remaining seven sectoral models contained two to four macro variables drawn from: GDP, industrial production, the unemployment rate, gross fixed capital equipment, the oil price, and the three-month real interest rate.
- Adjusted R squares ranged between 10 percent and 27 percent.
  - Note: Lower adjusted R squares attributed to high variance in quarterly series; similar annual-data models typically show adjusted R squares above 50 percent.

### Macroeconomic Variables included in Model-Selection (full list considered)
- Cyclical Indicators:
  - Gross domestic product (GDP)
  - Industrial Production International (IPI)
  - Dom. Industrial production less energy (IPexE)
- Price stability indicators:
  - Consumer price Index (CPI)
  - Inflation rate (InfR)
- Household Indicators:
  - Unemployment Rate (seasonal adjusted) (Uersa)
  - Private Consumption (privCon)
- Corporate Indicators:
  - Gross fixed capital formation (Gfctot)
  - Gross fixed capital equipment (GfcEq)
  - Gfctot/GDP
  - Ifo business climate index
- Interest rate indicators:
  - Nominal short-term interest rate
  - Real short-term interest rate
  - Nominal 5y term interest rate
  - Real 5y term interest rate
  - Nominal 10y term interest rate
  - Real 10y term interest rate
- External variables:
  - Exports
  - Export/GDP
  - Imports
  - FX-Rate USD
  - FX-Rate CHF
  - FX-Rate JPY
  - FX-Rate GBP
  - Oil price (Brent Crude) in EUR
- Equity:
  - Dow Jones Index

### B. CESE Credit Risk — Approach and limitations
- Reliable PD data generally not available for CESE; available time series too short to estimate econometric PD/NPL models.
- LLPs are used as the credit-risk measure for CESE because LLPs are the only credit-risk data reported from Austrian subsidiaries in CESE to the OeNB.
- Limitations of LLPs acknowledged (e.g., income smoothing); NPLs also have limitations (e.g., divergent regulations).
- Translation of macroeconomic stress test scenario to LLPs relies on expert judgment rather than econometric modeling due to short time series.

---

### APPENDIX II: MODELING INDIRECT CREDIT RISK INDUCED BY FOREIGN EXCHANGE RISK

### General methodology
- Indirect credit risk from exchange rate movements estimated via logarithmic changes of the loan loss provision ratio, ΔLLPR_l.
- A relative increase of the foreign exchange rate by Δr_FX increases outstanding euro volume of a foreign currency loan by Δr_FX ⋅ D, where D is outstanding volume in euros.
- Assumption (worst-case-boundary): economic loss from FX movement and repayment vehicle (RPV) performance is incurred instantaneously and must be fully covered in the current year, interpreted as a decrease of yearly income by Δr_I ⋅ I.
- Aggregate relation:
  - Δr_I = Δr_FX + Δr_RPV − Δr_D  (presented as (RPV FX) I D r r r Δ − Δ − = Δ)
- Logarithmic change of LLPR related linearly to logarithmic change of yearly income:
  - ΔLLPR_l = β̂ ⋅ ΔI_l

### Regression linking LLPR_l and GDP growth (proxy for income)
- Use real GDP-growth ΔGDP_l as proxy for ΔI_l for foreign currency loans across households and corporates.
- Estimated coefficient: β̂ = −5.2
- Regression statistics (Table 9):
  - Dependent Variable: LLPR_lΔ
  - Method: Least Squares
  - Sample: 1984 2007
  - Included observations: 23 after adjusting endpoints
  - Newey-West HAC Standard Errors & Covariance (lag truncation=2)
  - Constant: Coefficient 0.079155; Std. Error 0.040077; t-Statistic 1.975075; Prob. 0.0616
  - GDP_lΔ: Coefficient −2.528068; Std. Error 1.022124; t-Statistic −2.473348; Prob. 0.0220
  - R-squared 0.147836
  - Adjusted R-squared 0.107257
  - Log likelihood 31.18585
  - Durbin-Watson stat 0.569031
  - F-statistic 3.643137
  - Probability (F-statistic) 0.070063
  - Source: OeNB

### Leverage ratio and scaling parameters used
- Proxy for average size of a foreign currency loan D: approximately € 130,000.
- Total volume of foreign currency lending to domestic customers: € 49 billion from some 380,000 FX loans.
- Proxy for I (yearly income): real GDP per unit labor force ≈ €60,400 for 2007.
- Implied ID (I/D) estimate in the range between 2 and 2.5; chosen value ID = 5.2 (as used in the estimation).

### Formulae and capital impact calculation (preserving original expressions)
- Relationship between income change and LLPR:
  - (1) ΔLLPR = β̂ ⋅ ΔI
  - Using β̂ = −5.2 leads to specific algebra shown in text:
    - 1) 1(ˆ − + Δ = Δ β I LLPR r)
    - 1]1)(5.2[1]1)([ 5.2 ˆ −+Δ−Δ×−=−+Δ−Δ−=Δ − RPVFX RPVFX I D LLPR r r r r β
- Change in LLP assumed to directly and fully impact capital:
  - ΔCapital = − V ⋅ LLPR ⋅ {1]1)(5.2{[ −+Δ−Δ×− } (variable notation as in source)
  - V denotes a bank’s volume in loans in the respective currency.

---

*Source: _cr08204 - 65.      Other risks that require vigilance follow from the concentrated exposure of the*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2008/_cr08204.pdf_