## EXECUTIVE SUMMARY (content unit 1autea2020006) ## Source details **Canonical URL:** [EXECUTIVE SUMMARY (content unit 1autea2020006)](https://www.imf.org/-/media/files/publications/cr/2020/english/1autea2020006.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2020/english/1autea2020006.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2020/english/1autea2020006.pdf.json) --- ### Key structural features and vulnerabilities - Banking sector: "almost 600 banks" at unconsolidated level; "heavily tiered and clustered." - Complex ownership: "inverse ownership whereby small lower tier banks own large higher-tier banks" (notably Raiffeisen structure). - Institutional cooperation arrangements: institutional protection schemes (IPS), cross-guarantee schemes, and liquidity associations generating "significant financial interlinkages among institutions." - Profitability constraint: "structurally high cost-to -income ratios" despite consolidation. - CESEE exposure: remains "substantial" though lower than in the past; CESEE operations profitable but with "thin" interest margins and rising real estate exposures, including in Austria. ### Late-cycle macrofinancial metrics (selected) - GDP growth: "averaging 2.7 percent" in 2017–18. - Unemployment: "4.7 percent" in 2018. - Inflation: "1.9 percent" in 2018. - Growth outlook: "expected to ease to 1.6 percent in 2019, settling around the potential of 1.75 percent thereafter." - Private credit growth (2019Q1): households "4.2 percent" y/y; NFC "7.0 percent" y/y. - House prices: grew "around 7 percent over 2018"; overvalued "around 10-15 percent nationally, and by over 20 percent in Vienna." - Real estate transaction volumes: "13.3 percent in 2018." - Households’ debt: "50 percent of GDP in 2018Q4" (EA average "58 percent of GDP"). - Variable-rate mortgages in new loans: declined "from 83 percent to 44 percent" over the last four years. - FX mortgage loans: "9 percent of total outstanding loans" (driven by housing loans in Swiss francs). ### FSAP scope, scenarios, and stress-testing focus - Solvency adverse scenario: financial cycle downturn with "sharp foreign exchange (FX) market disruptions and sovereign stress" causing balance-sheet recession in Austria and CESEE and a sharp real estate price correction (20 percent peak-to-trough real estate price decline calibrated in adverse scenario). - Liquidity scenarios: large outflows from retail funding, wholesale funding, and both combined. - Contagion analysis: hypothetical contagious bank defaults including defaults within decentralized banking sectors; explicit testing of IPS resiliency and inward spillovers. ### Main solvency stress-test headline findings - Aggregate resilience: "Banks are, in aggregate, resilient to severe macrofinancial shocks." - Capital buffers: "sizeable relative to potential losses under stress." - Capital ratios: "would still be comfortably above minimum requirements in an adverse scenario, although most banks would fall into the conservation buffer range." - Drivers vary by institution: - OSIIs: "credit impairments and lower net interest income are key factors driving capital depletion." - Smaller non-OSIIs: impact "mainly attributed to equity participation structures" (participation/inverse ownership channel). - Inverse ownership: under favorable conditions contributes positively to capital generation via profit redistribution; under unfavorable conditions "can become a shock amplifier." --- ### The financial system structure and CESEE exposures ### Size and composition - Banking sector assets: about 250 percent of GDP (EUR986 billion) at end-2018. - Banking assets share: about 75 percent of total financial system assets (EUR1,314 billion) at end-2018. - Asset shares: mutual funds 14 percent, insurance firms 10 percent, pension funds 2 percent of financial system assets. - Financial system exposure to CESEE: 22 percent of banking system assets. - Austria’s share of CESEE claims by EU-15 banks: 22 percent. - Banks’ reliance on region for profits: 42 percent of total. ### Sector structure and IPS arrangements - Unconsolidated institutions: almost 600 (down from nearly 1,000 two decades ago). - LSI sector assets: 32 percent of total banking system assets. - Tiered/cooperative segments account for half of banking system assets. - Raiffeisen sector structure: - Three-layer: 386 local banks (first-tier) → eight regional banks (RLBs, second-tier) → RBI (third-tier). - RLBs hold around 60 percent of RBI ("inverse ownership"). - Six regional IPSs and a federal IPS (RBI and RLBs members). - Sparkassen segment: - 49 credit institutions; central institution Erste Group Bank AG (Erste); IPS forming cross-guarantee scheme; Sparkassen-IPS recognized as DGS. - Erste controls five savings banks by majority ownership and 43 banks via cross-guarantee; internationally active in CESEE. - Volksbanken segment: eight regional banks; unlimited cross-guarantee liability association operating with centralized risk management. ### IPS funding rules and regulatory implications - Ex-ante fund: contributions benchmarked annually against stress test results and confirmed by regulator. - Ex-post contributions: Risk Council may require ex-post payments up to maximum of 50 percent of the sum of the last three years average operating results; if insufficient, additional payments may be required. - Decision default: if Risk Council does not decide within two weeks, members must contribute at maximum 25 percent of total capital in excess of regulatory requirements. - Contribution caps: each contribution limited by member’s own capital requirements plus a cushion of 10 percent. - Regulatory exemptions: 0 percent risk weights on certain exposures (Article 113.7 CRR) and non-deduction of participations in central institution (49.3 CRR); exemptions withdrawn upon member insolvency or resolution. --- ### Recent banking performance, credit quality, and vulnerabilities ### Capitalization and asset quality - CET1 ratio: rose from 11.6 percent in 2013 to 15.4 percent in 2018; increase subsided recently partly due to higher dividend payouts. - CET1 by size: LSI institutions CET1: 17.0 percent; SIs CET1: 13.6 percent. - NPLs: fell from 8.6 percent in 2013 to 2.6 percent in 2018. - NPLs in CESEE subsidiaries: fell from 14.0 percent in 2013 to 3.2 percent in 2018. ### Funding, profitability, and costs - Customer deposits: ~40 percent of liabilities. - Debt securities issuance: ~14 percent of total liability outflows. - CTI ratio: around 65 percent in 2018 (structurally high); Austrian SIs posted CTIs slightly below EA average. - Consolidated net interest margin: 1.0 percent; CESEE margins: around 1.6 percent. - Intragr0up liquidity transfers to CESEE subsidiaries: declined from 15 percent of subsidiaries' assets in 2011 to below 9 percent in 2018. - CESEE subsidiaries CTI: 51 percent (lower than consolidated but rising). ### Key vulnerabilities - Structural: large, interconnected system; clustered CESEE exposures; low domestic profitability due to margin compression and overcapacity. - Cross-border spillovers: shocks in central institutions or foreign subsidiaries can propagate via participations, IPS, cross-guarantees, liability associations, and DGS. - FX exposure: FC loan stock declined but remains high and susceptible to home currency depreciation. - Real estate exposures in CESEE (Czech Republic, Slovakia) increased amid market vulnerabilities. - Profitability pressure drivers: low-for-long rates encouraging fixed rates; possible amplified margin shocks in downturn; international competition and market maturation. --- ### Solvency stress-testing design, modelling, and main quantitative findings ### Coverage, horizon, and methodology - Coverage: macroprudential top-down stress test covering 440 banks (>95 percent of banking system assets); supervisory data as of end-2018. - Solvency test: top-down, balance-sheet approach; Q4 2018 cut-off; two scenarios (baseline and adverse); horizon three years (2019–2021); static balance sheet assumption. - Risks covered: credit risk, market risk, net interest income, non-interest income, and participation risk (equity participations). - Tools: OeNB’s ARNIE infrastructure with IMF satellite model inputs. - Accounting/regulatory treatment: assumed accounting impairments match regulatory expected losses due to mixed reporting (IFRS 9 and national GAAP). ### Adverse scenario calibration (selected) - Adverse features: - Financial cycle downturn with sharp FX market disruptions and sovereign stress. - Implies deviation of Austrian real GDP from baseline of 6.9 percent by 2021. - 2.3 standard deviation move in two-year cumulative real GDP growth rate. - 20 percent peak-to-trough decline in real estate prices. - Output shocks in CESEE range between 8.1 and 13.6 percent deviation from baseline. - At -2.4 percent GDP contraction over first two years, more severe than global financial crisis (-2.0 percent over 2009–10). ### Key stress-test quantitative results - Aggregate CET1 ratio evolution: - Baseline: aggregate CET1 capital ratio would increase by 2.3 percentage points by 2021. - Adverse: aggregate CET1 capital ratio would decline by 4.4 percentage points to 11.1 percent by 2021. - Bank-level shortfalls: - Five small banks (out of 440) would have ending CET1 below minimum CET1 requirement in the adverse scenario. - Total capital shortfall ≈ EUR 5 million (0.01 percent of aggregate system CET1 capital). - Primary drivers of capital depletion in adverse scenario: - Credit impairments: 4.9 percentage points of CET1 depletion under adverse scenario (versus 1 percentage point decline in baseline). - Net interest income: on average lower by almost 15 percent for the horizon in the adverse scenario; NII reduction measured at 18.9 percent relative to 2018 in adverse scenario in one reported estimate. - Increase in RWAs due to credit quality deterioration: additional CET1 depletion of 0.7 percentage points (impact concentrated in OSIIs). - FC loans add-on: 0.6 percentage points of CET1 depletion across all countries due to additional impairment. - Participation/inverse ownership channel: - Non-OSIIs participation valuation contribution: +1.8 percentage points in baseline versus -1.9 percentage points in adverse. - Participation channel contribution to system capital generation capacity: 0.8 percentage points. ### Heterogeneity: OSIIs vs non-OSIIs - OSIIs: more severely hit on NII—OSII group NII reduction in adverse scenario ≈ 20.1 percent average over three years. - Non-OSIIs: NII reduction in adverse scenario ≈ 6 percent. - Main OSII NII driver: exposure to non-EU countries with sharper reference rate increases and volatile reference rates. --- ### Real estate, household vulnerabilities, and mortgage modelling ### Household and mortgage exposures - Household debt: 87 percent of disposable income. - House price overvaluation: inflates collateral values backing new mortgages. - Significant share of new mortgages non-compliant with recommended DSTI limit or minimum down-payment. - Real estate exposures increased across financial sector; systemic risk build-up noted. ### Adverse scenario specification (real estate and household) - Adverse scenario includes: - (i) sharp global weakening and prolonged Austrian slowdown; - (ii) sharp increase in risk premia and higher debt servicing costs; - (iii) balance-sheet recession in CESEE with large FX moves and asset repricing; - (iv) sharp correction in Austrian real estate prices (20 percent peak-to-trough). - Structural mortgage model: - Structural DSR/LTV joint-distribution approach using TUI model adaptations. - Defaults require "double trigger": borrower inability to service and inability to prepay via collateral sale (post-shock LTV > 1). - Behavioral survival rule: linear survival rule between wealth buffer of 8 and 36 months. - Monte Carlo simulation to produce 3-year loss rates by DSR/LTV partition. ### Model outputs (selected 3-year loss rates, Table IV.1) - Baseline 3-year total loss rates (Total 3 Year): - Austria: 0.39 - Czech Republic: 0.92 - Slovakia: 0.08 - Hungary: 0.30 - Romania: 0.08 - Bulgaria: 0.19 - Slovenia: 0.05 - Croatia: 0.34 - Poland: 0.16 - Russia: 0.29 - Ukraine: 0.02 - Germany: 0.09 - EA/ROW: 0.09 - UK: 0.23 - Spain: 0.19 - France: 0.14 - US: 0.06 - Turkey: 0.51 - Adverse 3-year total loss rates (Total 3 Year): - Austria: 1.38 - Czech Republic: 1.44 - Slovakia: 1.43 - Hungary: 5.10 - Romania: 2.67 - Bulgaria: 3.48 - Slovenia: 0.87 - Croatia: 4.47 - Poland: 1.58 - Russia: 2.21 - Ukraine: 1.40 - Germany: 0.57 - EA/ROW: 1.26 - UK: 1.10 - Spain: 1.61 - France: 1.92 - US: 1.49 - Turkey: 2.11 - Interpretation: projected loss rates relatively moderate for most countries; higher adverse losses associated with strong positive interest rate shocks (e.g., Hungary, Romania, Bulgaria, Croatia). ### Foreign currency (FC) and repayment vehicle (RV) loans - FC/RV module: quantifies losses from FC currency appreciations and RV underperformance; covers domestic RV loans in EUR, CHF, JPY and CESEE CHF, EUR (in non-EUR countries), USD exposures. - Adverse FX assumptions: CHF appreciates versus EUR by around 50 percent; EUR appreciates versus local CESEE currencies in the range of 10 to 30 percent. - Domestic model relationship example: for linear slope β = -2.5, ΔLLPR = 2.5 * D/I * ΔFX; debtor leverage D/I set to 2.5 per supervisory minimum. --- ### Liquidity and funding resilience — LCR, cash-flow, NSFR, and IPS effects ### LCR-based stress tests: scenarios and key metrics - Regulatory LCR hurdle: 100 percent. - Asset-weighted average LCR: 164.7 percent. - Outflows-weighted average LCR: 148 percent. - Asset-weighted LCR under scenarios: - Retail funding outflows: 128 percent. - Wholesale funding outflows: 119.5 percent. - Severe combined scenario (including market price shock): 104.3 percent. - EUR LCR: 145.8 percent (regulatory) → 96.9 percent (severe scenario). - USD LCR: starting system-wide USD LCR average 56.6 percent; severe scenario USD LCR 44.1 percent; USD outflows in severe scenario amount to 6.1 billion. ### Cash-flow (3-month) analysis: design and outcomes - Horizon: 3 months; contractual maturity ladder and counterbalancing capacity measured after scenario haircuts. - Assumption: banks continue to provide credit; roll-over rates of maturing retail and corporate loans = 100 percent to increase severity. - Aggregate net liquidity position (counterbalancing capacity minus net outflows): - Initial level: 21 percent of system assets. - Severe scenario: 8.8 percent of system assets. - Combined liquidity shortfall in extreme (severe) case with IPS support: EUR 1.5 billion. - No-IPS scenario combined shortfall: EUR 1.7 billion. - Conclusion: IPS enhances resilience; banking system can withstand prolonged adverse liquidity conditions even without IPS support; some regional IPSs show shortfalls in severe scenario but federal IPS layer buffers are adequate. ### NSFR exploratory stress test - Proxy NSFR: asset-weighted aggregate NSFR 122 percent. - System-wide NSFR under stress: 112 percent. - Share of system assets with NSFR < 100 percent: - Initial: 2.2 percent. - Under stress: >28 percent. - Note: NSFR to come into force in 2021Q2. ### Liquidity drivers and funding profile - Retail and corporate deposits: almost 40 percent of liabilities. - Repo segment: does not play a significant role. - Derivatives-related liabilities: ~18 percent of total funding (important due to CESEE presence). - Asset and liability derivatives: mostly linked FX and cross-currency swaps and seem balanced. --- ### Interconnectedness, contagion analysis, and network results ### Network analysis approach and parameters - Methods: Furfine algorithm, Espinosa-Sole tool, CoMap methodology for cross-border contagion. - Data: consolidated gross exposures from central credit registry (credit instruments above EUR 350,000); exposures to affiliated banks within cooperative sector deducted for some exercises. - Assumed contagion parameters: - λ1 = 0.550 percent LGD on unsecured lending (presented as 0.5 in annex summary elsewhere). - λ2 = 0.2 20 percent LGD on secured lending. - ρ = 0.35 share of lost funding non-replaceable. - δ = 0.3 30 percent discount on asset sales. - 65 percent roll-over ratio of interbank liabilities; 80 percent recovery rate for secured interbank lending; 50 percent recovery for unsecured; 30 percent fire-sale discount. - Failure criterion: CET1 capital falls below 4.5 percent. ### Contagion findings and sensitivity to cooperative sector treatment - Results sensitive to inclusion/exclusion of cooperative sector affiliated exposures: - Including gross exposures (affiliated entities): - Maximum contagion index for OSII group: 41 percent of total capital (unweighted). - Median bank contagion for OSII (gross): 1.1 percent. - Excluding affiliated exposures: - Maximum contagion ≈ 1.2 percent of system capital. - Median contagion ≈ 0.5 percent. - CoMap cross-border contagion: - Median contagion index from Austrian OSIIs to Austrian banking system: 0.83 percent. - Contagion at 10th percentile: 0.5 percent; at 90th percentile: 2.1 percent. - Median contagion from Austrian OSIIs to EA: 0.12 percent capital depletion. - Outward/inward spillovers between Austrian banks and CESEE/EA limited; inward spillovers from CESEE and Italy negligible. - Caveat: analyses often exclude equity exposures which would amplify losses if included. --- ### Policy recommendations, stress-testing enhancements, and monitoring priorities ### Key policy recommendations (selected) - General systemic monitoring and buffers: - "Enhance monitoring capacity and oversight of inward spillover risks due to the inverse ownership structure of the Raiffeisen sector." Priority: H. Timeframe: I. - "Perform a comprehensive assessment of the systemic importance of the central institution in the decentralized sectors and adjust capital buffers appropriately to offset systemic vulnerabilities linked to the tiered structure." Priority: H. Timeframe: I. - "Continue to monitor the process of improving structural profitability." Priority: H. Timeframe: NT. - "Enhance data collection (especially with respect to risk parameters) to better cover the CESEE region and develop more granular approaches in analyzing financial stability risks coming from such exposures." Priority: M. Timeframe: MT. - Solvency stress testing enhancements: - "Enhance the stress testing framework to account for second round effects, dynamic balance sheets, and contagion/spillover effects..." Priority: M. Timeframe: MT. - "Ensure resources and organizational structure are adequate given complexity..." Priority: M. Timeframe: NT. - "Further develop the macroprudential and top-down perspective... full top-down satellite models complementary to EBA methodology." Priority: H. Timeframe: NT. - "Enhance capacity to calibrate tailored scenarios targeting Austrian banking sector cyclical and structural vulnerabilities." Priority: H. Timeframe: MT. - Liquidity stress testing and IPS integration: - "Further develop systemic liquidity stress test components and enhance their macroprudential angle." Priority: L. Timeframe: NT. - "Enhance forward-looking aspects of cash-flow liquidity analysis by carefully assessing cliff effects over time bands." Priority: M. Timeframe: NT. - "Fully integrate the analysis on the liquidity support buffers due to the IPS structures in the standard liquidity analysis." Priority: H. Timeframe: MT. - Interconnectedness and monitoring: - "Conduct network analysis on a detailed breakdown of bilateral exposures by instrument, identifying the liquidity reserve held at the central institution and adding equity exposures in the cooperative sector." Priority: L. Timeframe: MT. - "Monitor interconnectedness on a quarterly basis." Priority: L. Timeframe: MT. ### Operational recommendations for stress-testing function and data - Strengthen OeNB ARNIE infrastructure use and ensure organizational resources match the complexity of stress testing across 440 banks. - Integrate sensitivity, contagion, and interconnectedness tools into standard macroprudential stress-testing toolkit. - Increase modelling granularity to capture geographical dispersion and risk diversity; integrate satellite/core models as alternatives to standard EBA approach. - Prioritize improving CESEE exposure data granularity via cooperation across authorities and entities. --- ### Modelling infrastructure and parameter calibration notes ### Satellite models and PD/LGD approach - PD PiT models: Bayesian Model Averaging (BMA) for corporate and financial segments across 18 geographies (AT, CZ, SK, HU, RO, BG, SI, HR, PL, RU, UA, DE, UK, ES, FR, US, TR, ROW). - Logit transformation applied; models include autoregressive lags and macro-financial regressors (real GDP, investment, consumption, inflation, interest rates, credit, stock indices, unemployment). - LGD PiT: OeNB internal approach; effective LGD = LGD_unsec × (1 - collateral ratio); collateral shocked by haircuts; LGD_unsec calibrated to CCR data for Austria and World Bank Doing Business anchors for others. - Mortgage structural model: TUI-based, DSR/LTV joint distribution, behavioral survival rule (8–36 months wealth buffer), Monte Carlo to produce 3-year loss rates. ### Market risk and NII treatment - Market risk: valuation losses via modified duration for interest-rate sensitive assets and spread shocks; FVTPL losses to P&L, FVOCI losses via OCI; hedges assumed ineffective. - NII: semi-structural re-pricing model following EBA methodology; assets/liabilities mapped to COREP/FINREP segments; pass-through constraints applied; maturing instruments replaced at scenario-implied reference and margin rates. - Non-interest income: constrained by ARNIE implementation; reductions applied (15 percent reduction on NFCI and NTI; 25 percent reduction on dividend income and non-consolidated profit shares); administrative expenses floored to cut-off levels. --- *Source: EXECUTIVE SUMMARY and selected chapters and annexes, FSAP Technical Note and accompanying material (content unit 1autea2020006).* ### EXECUTIVE SUMMARY __________________________________________________________________________ 6 ### EXECUTIVE SUMMARY ### Key structural features and vulnerabilities - The banking sector counts "almost 600 banks" at an unconsolidated level and is "heavily tiered and clustered." - Complex ownership structures include "inverse ownership whereby small lower tier banks own large higher-tier banks" and institutional cooperation arrangements: institutional protection schemes (IPS), cross-guarantee schemes, and liquidity associations. - These features generate "significant financial interlinkages among institutions." - Banks’ operating profits are constrained by "structurally high cost-to -income ratios" despite some consolidation. - Exposure to Central, Eastern, and Southeastern Europe (CESEE) markets remains "substantial" though lower than in the past; CESEE operations have been profitable but have "thin" interest margins and rising exposure to booming real estate sectors, including in Austria. ### Late-cycle macrofinancial environment (selected metrics) - Austria’s GDP growth: "averaging 2.7 percent" in 2017–18. - Unemployment: "4.7 percent" in 2018. - Inflation: "1.9 percent" in 2018. - Growth outlook: "expected to ease to 1.6 percent in 2019, settling around the potential of 1.75 percent thereafter." - Private credit growth (2019Q1): households "4.2 percent" year-on-year; nonfinancial corporate (NFC) sector "7.0 percent" year-on-year. - House prices: grew "around 7 percent over 2018"; estimated overvalued "around 10-15 percent nationally, and by over 20 percent in Vienna." - Real estate transaction volumes: growth "13.3 percent in 2018." - Households’ debt: "50 percent of GDP in 2018Q4" (EA average "58 percent of GDP"). - Variable-rate mortgages in new loans: declined "from 83 percent to 44 percent" over the last four years. - Foreign currency (FX) mortgage loans: "9 percent of total outstanding loans" (driven by housing loans in Swiss francs). ### Scope and scenarios of the FSAP analysis - Solvency stress test adverse scenario: a financial cycle downturn with "sharp foreign exchange (FX) market disruptions and sovereign stress" producing a balance-sheet recession in Austria and the CESEE region and a sharp real estate price correction. - Liquidity stress test scenarios: considered "large outflows from the retail and wholesale funding segments—and the two combined." - Contagion analysis: assessed "hypothetical contagious bank defaults including such defaults within the decentralized banking sectors." - The resilience of IPS structures and their ability to provide capital and liquidity support to affiliated members under stress was explicitly tested. ### Main solvency stress-test findings - "Banks are, in aggregate, resilient to severe macrofinancial shocks." - Capital buffers are "sizeable relative to potential losses under stress." - "Capital ratios would still be comfortably above minimum requirements in an adverse scenario, although most banks would fall into the conservation buffer range." - Impact drivers differ across institution types: - For Other Systemically Important Institutions (OSIIs): "credit impairments and lower net interest income are key factors driving capital depletion." - For the majority of smaller non-OSIIs: impact is "mainly attributed to equity participation structures that form the backbone of the heavily tiered Austrian banking system." - The FSAP notes that "under favorable economic conditions inverse ownership contributes strongly to their capital generation by allowing partial redistribution of profits higher tier banks in the Raiffeisen sector earn on their more profitable international business." - Conversely, "under unfavorable economic conditions, inverse ownership can become a shock amplifier by propagating losses through the downward revaluation of equity participations, adding to profitability pressures in domestic market operations." ### Liquidity and funding resilience - "Banks hold sufficient liquidity buffers to withstand sizeable funding outflows." - The system is resilient due to "strong counterbalancing capacity and sizable deposit funding base." - Liquidity associations in the tiered sector allow "pooling of liquidity among members, which contributes to overall financial stability." ### Interconnectedness and contagion risks - "High banking system interconnectedness warrants close monitoring." - Authorities have imposed OSII buffers also at the unconsolidated level to target vulnerabilities related to interconnectedness. - Additional monitoring recommended, including: - compiling detailed breakdowns of exposures within the decentralized sectors to better gauge contagion potential; - deploying top-down reverse stress tests to assess IPS resiliency; and - setting up a permanent data monitoring framework and toolkit for assessing inward spillovers and contagion effects. ### Institutional-systemic context (selected items) - There are "seven OSII banks on a consolidated level": Erste Group Bank, Raiffeisen Bank International, UniCredit Bank Austria, Raiffeisenlandesbank Oberoesterreich, BAWAG P.S.K, Raiffeisen-Holding Niederoesterreich-Wien, and Volksbanken Verbund. - The Technical Note was prepared by Dimitrios Laliotis (MCM) and Lucyna Gornicka (MCM), Izabela Karpowicz (EUR) and Sohaib Said (MCD) under the guidance of Laura Valderrama (MCM). ### Key policy recommendations (from Table 1) - General - "Enhance monitoring capacity and oversight of inward spillover risks due to the inverse ownership structure of the Raiffeisen sector." Priority: H. Timeframe: I. - "Perform a comprehensive assessment of the systemic importance of the central institution in the decentralized sectors and adjust capital buffers appropriately to offset systemic vulnerabilities linked to the tiered structure." Priority: H. Timeframe: I. - "Continue to monitor the process of improving structural profitability." Priority: H. Timeframe: NT. - "Enhance data collection (especially with respect to risk parameters) to better cover the CESEE region and develop more granular approaches in analyzing financial stability risks coming from such exposures." Priority: M. Timeframe: MT. - Solvency Stress Testing - "Enhance the stress testing framework to account for second round effects, dynamic balance sheets, and contagion/spillover effects as an effective tool to assess vulnerabilities related to the complexity of the system." Priority: M. Timeframe: MT. - "Ensure that resources and the organizational structure are adequate to meet the objective of the stress testing framework given the complexity of the Austrian banking system and responsibilities related to supervisory stress testing and European coordination." Priority: M. Timeframe: NT. - "Further develop the macroprudential and top-down perspective of the stress test infrastructure by developing full top-down satellite models that are complementary to the ones relying on the European Banking Authority (EBA) methodology." Priority: H. Timeframe: NT. - "Enhance the capacity to calibrate tailored scenarios which target the Austrian banking sector’s cyclical and structural vulnerabilities that are complementary to the ones used for microprudential purposes." Priority: H. Timeframe: MT. - Liquidity Stress Testing - "Further develop the systemic liquidity stress test components end enhance their macroprudential angle." Priority: L. Timeframe: NT. - "Enhance the forward-looking aspects of the cash-flow liquidity analysis toolkit by carefully assessing cliff effects over time bands." Priority: M. Timeframe: NT. - "Fully integrate the analysis on the liquidity support buffers due to the IPS structures in the standard liquidity analysis." Priority: H. Timeframe: MT. - Interconnectedness Analysis - "Conduct network analysis on a detailed breakdown of bilateral exposures by instrument, identifying the liquidity reserve held at the central institution and adding equity exposures in the cooperative sector." Priority: L. Timeframe: MT. - "Monitor interconnectedness on a quarterly basis." Priority: L. Timeframe: MT. *Source: EXECUTIVE SUMMARY, FSAP Technical Note and accompanying material.* ### 5. The financial system is dominated by banks and is heavily exposed to CESEE countries. ### 5. The financial system is dominated by banks and is heavily exposed to CESEE countries. ### Financial system structure and exposures - Banking sector assets amounted to about 250 percent of GDP (EUR986 billion) at end-2018. - Banking assets represented about 75 percent of total financial system assets (EUR1,314 billion) at end-2018. - Mutual funds, insurance firms, and pension funds account for 14 percent, 10 percent, and 2 percent of financial system assets, respectively. - Financial system exposure to CESEE countries remains high at 22 percent of banking system assets, despite declines after a major bank restructuring in 2016. - Austria accounts for 22 percent of CESEE related claims of EU-15 banks. - Banks’ reliance on the region for profit generation is at 42 percent of total. ### Banking sector composition and corporate structure - The banking sector comprises almost 600 unconsolidated institutions, down from nearly a thousand two decades ago. - The LSI sector has total assets amounting to 32 percent of total banking system assets. - The banking system can be divided into categories: joint stock banks, cooperative banks, savings banks, regional banks, and other institutions. - Banks with tiered corporate structures and institutional cooperation arrangements account for half of banking system assets. - Cooperation arrangements include liquidity associations, cross-guarantee schemes, and institutional protection schemes (IPS). - IPS features: - Legal obligation to grant solvency and liquidity support to participating members. - Subject to comprehensive supervision due to extended aggregation levels. - Required to establish an ex-ante fund and subject to external audit. - Financial support may take the form of loans, guarantees, or capital and liquidity injections. - Governed by a Risk Council that decides measures if an institution triggers early warning indicators. - IPS must carry out stress tests at regular intervals and take preventive action before resolution measures. ### Key segment structures - Raiffeisen sector: - Three-layer structure: 386 local banks (first-tier) hold shares in one of eight regional banks (RLBs, second-tier); RLBs form liquidity associations with the third-tier bank, RBI. - RLBs hold around 60 percent of RBI ("inverse ownership"). - RBI is a listed bank with several foreign subsidiaries in CESEE, including Russia. - Six “regional” IPSs exist between local banks and their RLBs; there is a federal IPS with RBI and the RLBs as members. - Sparkassen segment: - 49 credit institutions including central institution Erste Group Bank AG (Erste). - Created a banking group and an IPS forming a cross-guarantee scheme, ex lege according to CRR. - Sparkassen-IPS recognized as a deposit guarantee scheme (DGS). - Erste controls five savings banks through majority ownership, and 43 banks through the cross-guarantee scheme. - Erste is internationally active in CESEE and has 70 percent of free float equity. - Volksbanken (VB) segment: - Features eight regional banks, including VB Wien AG and one specialized institution. - Formed a liability association with unlimited cross-guarantee scheme, operating increasingly as a single institution with uniform risk management and centralized processes. ### IPS funding and regulatory implications - IPS funding and contribution rules: - Each Raiffeisen IPS member must contribute to an ex-ante fund; the amount is benchmarked annually against stress test results and confirmed by the regulator. - If ex-ante fund is insufficient, members are required to provide ex-post contributions. - The Risk Council may require ex-post payments up to a maximum of 50 percent of the sum of the last three years average operating results of all members; if insufficient, it can require additional payments. - If the Risk Council does not decide within two weeks about further payments, it must decide that members contribute at the maximum 25 percent of their total capital in excess of regulatory requirements. - Each contribution (ex-ante or ex-post) is limited by the member’s own capital requirements plus a cushion of 10 percent. - Regulatory exemptions linked to contractual arrangements: - Include 0 percent risk weights on exposures and elimination of multiple gearing of capital at institution level. - Authorized privileges for IPS members are reflected in Article 113.7 CRR (0 percent risk weight) and 49.3 CRR (non-deduction of participations in central institution). - Exemptions are automatically withdrawn if a member becomes insolvent or enters resolution because not all available support is provided. ### Recent banking sector performance and credit quality - Capitalization and credit quality trends: - CET1 ratio rose from 11.6 percent in 2013 to 15.4 percent in 2018; the increase subsided in the last two years, partly due to higher dividend payouts. - LSI institutions CET1: 17.0 percent; SIs CET1: 13.6 percent. - Nonperforming loans (NPLs) shrank from 8.6 percent in 2013 to 2.6 percent in 2018. - NPLs in subsidiaries in CESEE declined from 14.0 percent in 2013 to 3.2 percent in 2018. - Funding and liabilities: - Customer deposits at around 40 percent of liabilities underpin banking system stability. - Debt securities issuance accounts for around 14 percent of total liability outflows and is mostly used by larger banks. - Profitability and costs: - Domestic profitability is weak but consolidated capital ratios improved due to enhanced net interest margins on international exposures and reduction of credit impairments. - Operating profits are affected by a structurally high cost-to-income (CTI) ratio at around 65 percent in 2018. - Austrian SIs posted CTIs slightly below the EA average. - Internal income generation remains modest and mainly driven by CESEE profitability. - CESEE subsidiary performance: - In 2018 Austrian subsidiaries in CESEE saw above average profitability, but performance has likely peaked. - Net interest margin at consolidated level: 1.0 percent; margins in CESEE region: around 1.6 percent. - Share of intragroup liquidity transfers between Austrian parent banks and CESEE subsidiaries declined from 15 percent of subsidiaries' assets in 2011 to below 9 percent in 2018. - CESEE subsidiaries’ CTI ratios are lower than at the consolidated level (51 percent) but have been rising. ### Risks and vulnerabilities - Structural vulnerabilities: - Large and interconnected banking sector. - Clustered CESEE exposures across the largest banks and insurers. - Structurally low domestic profitability due to margin compression, overcapacity, and weak cost efficiency. - Cross-border and spillover risks: - Stress in the central institution of decentralized sectors or in Austrian foreign subsidiaries can spread into the domestic system because of equity participations, IPS, cross-guarantee schemes, liability associations, and DGS. - CESEE region vulnerabilities: financial market volatility, capital outflows, and foreign exchange (FX) swings. - Stock of foreign currency (FC) loans has declined but remains high and susceptible to home currency depreciation. - Subsidiaries’ real estate exposures—especially in Czech Republic and Slovakia—have increased amid rising real estate market vulnerabilities. - Profitability pressures ahead: - Ability to maintain solid net interest margins may be challenged by: - Low-for-long interest rate environment encouraging borrowers to opt for fixed rates. - In an economic downturn, shocks to margins on foreign operations could be amplified by elevated costs of risk, impairing internal capital generation. - Rising competition and market maturation effects internationally. ### FSAP analysis scope and stress-testing approach - Coverage: - The FSAP team, with OeNB, performed a macroprudential stress test covering 440 banks, accounting for over 95 percent of banking system’s total assets. - Supervisory data as of end-2018 were used. - Solvency stress test: - Top-down exercise using a balance sheet approach with regulatory and supervisory data; Q4 2018 cut-off date. - Two scenarios: baseline (based on July 2019 WEO forecast) and adverse. - Horizon: three years (2019 to 2021) under a static balance sheet assumption. - Included sensitivity tests; OeNB’s ARNIE infrastructure used with inputs estimated from FSAP satellite models. - Liquidity stress test: - Top-down exercise using regulatory data. - Assessed liquidity risks via Liquidity Coverage Ratio (LCR) tests, cash-flow analysis, and NSFR stress tests. - LCR tests conducted in total and by significant currencies under three scenarios: large outflows from i) retail funding, ii) wholesale funding, or iii) both combined. - Cash-flow analysis assessed banks’ counterbalancing capacity under varying severity. - NSFR stress tests were exploratory (NSFR not yet a binding minimum requirement). - Network analysis: - Assessed how distress in an individual institution could impair financial stability through default cascades in the interbank market. - Included credit shock simulation (counterparty default erodes lender’s capital) and funding shock simulation (funding counterparty default induces liquidity shortfall and fire-sale losses). *Source: IMF staff, Austria FSAP chapter — "The financial system is dominated by banks and is heavily exposed to CESEE countries."* ### 16. Vulnerabilities are building up in residential real estate. Although household credit has ### 16. Vulnerabilities are building up in residential real estate. Although household credit has ### Key findings on real estate and household vulnerabilities - Household debt stands at 87 percent of disposable income. - House price overvaluation inflates collateral values underlying new mortgages. - Low borrowing costs support households' demand for loans and debt servicing capacity. - A considerable share of new mortgages does not comply with the recommended DSTI limit or the minimum down-payment. - Real estate exposures have increased across various parts of the Austrian financial sector. - Given the increasing importance of the real estate market for financial stability, these trends point to a build-up of systemic risk. ### Macrofinancial scenarios (stress-test narrative) - The adverse scenario includes four systemic risk sources: - (i) a sharp weakening in the global outlook resulting in a prolonged growth slowdown in Austria; - (ii) a sharp increase in risk premia due to an abrupt deterioration in market sentiment that would increase debt servicing costs for firms and households and trigger asset repricing; - (iii) a balance sheet recession in CESEE countries amplified by large FX moves, lower asset price valuations, and worsen credit risk performance; and, - (iv) a sharp correction in real estate prices in Austria, resulting in higher impairment charges for banks, caused by defaults or delayed loan repayments. - Scenario calibration highlights: - Baseline scenario expectations: - GDP growth to slow to 1.6 percent in 2019 and settle around potential growth of 1.75 percent in the medium term. - Domestic demand growth to moderate from over 2.50 percent to around 1.25 percent in 2019. - Inflation to decelerate below 2 percent. - Adverse stress scenario features: - Calibrated using the Global Macrofinancial Model (GFM). - Financial cycle downturn with sharp FX market disruptions and sovereign stress, generating a balance-sheet recession in Austria and CESEE. - Implies a deviation of Austrian real GDP growth from its baseline of 6.9 percent by 2021. - A 2.3 standard deviation move in two-year cumulative real GDP growth rate. - A 20 percent peak-to -trough decline in real estate prices. - Output shocks in the CESEE region range between 8.1 and 13.6 percent deviation from baseline. - At -2.4 percent GDP contraction over the first two years of stress, the adverse scenario is more severe than the global financial crisis (-2.0 percent GDP contraction over 2009–10) and the 2018 EA FSAP scenario for Austria (-2.0 percent GDP contraction over 2018–19). ### Solvency stress-test design and implementation - Two scenarios calibrated over a three-year horizon to assess Austrian banks' resilience to severe macrofinancial conditions. - The FSAP top-down solvency stress test covered: - Credit risk (credit impairments; impact on RWAs, funding, and interest rate risk). - Market risk (repricing and spread risks for interest rate sensitive assets; equity and commodity risks). - Net interest income and non-interest income risks. - The FSAP team used the OeNB’s ARNIE infrastructure to run the solvency ST exercise, importing IMF satellite model outputs as parameters into ARNIE. - Balance sheet and income projection assumptions: - Static balance sheet assumption using Q4 2018 cut-off date asset and liability composition. - Sum of performing and nonperforming exposures kept constant at cut-off levels. - No increases in capital or managerial actions assumed. - Accounting and regulatory treatment: - Exercise assumed accounting impairments match regulatory expected losses due to mixed reporting (IFRS 9 and national GAAP) and ARNIE constraints. - RWAs projection accounted for defaulted credit formation and IRB portfolio quality deterioration; RWAs kept constant for STA portfolios and for operational and market risk at cut-off levels. - Income projections included net interest income, non-interest income and expenses, trading income, FVTPL portfolios, credit loss provisions, and tax charges. - Hurdle rates and capital treatment: - Stress test benchmarked against regulatory minimum and target values for CET1, T1, and total capital, including a capital conservation buffer (CCB) of 2.5 percent and a Systemic Risk Buffer (SyRB) for OSIIs. - Under the baseline scenario banks expected to maintain capital ratios above total target requirements (minimum, CCB, SyRB, and P2R). - In the adverse scenario preserving the CCB and the SyRB was not required. - Tax rates and dividend payout ratios: - Under both scenarios, tax rates and dividend pay-out ratio set at 30 percent. - If a bank made losses during a year, a zero dividend was applied. ### Credit risk analysis and modelling approach - Credit risk is the most important risk factor for the Austrian banking system due to the dominance of credit exposures in total assets. - International exposure feature: - CESEE exposures correspond to 32.3 percent of total exposures for OSII banks and 6.8 percent of total exposures for non-OSII banks. - Segmentation and models: - SIs: exposures allocated to corporates, financials, retail mortgages, retail consumer, sovereign, and sub-sovereign with satellite models estimated for all segments across material geographies. - LSIs: different segmentation leveraging granular Austrian Credit Register data; mapping to SI segments allowed use of common satellite forward paths. - The FSAP used OeNB LGD PiT scenario projections and OeNB estimated starting points for credit parameters. - PD projections for corporate and financial asset classes used satellite econometric models informed by Moody’s CreditEdge historical default rates; Bayesian Model Averaging (BMA) used in PD PiT model estimation. - Mortgage and real estate-related retail portfolios projected using a structural model driven by borrower affordability metrics (DSR) and house price shocks; model assumes a “double trigger” of default (borrower cannot service loan and cannot sell collateral to prepay loan). - Retail and other segments proxied using multipliers of corporate and mortgage PD paths due to low materiality and lack of historical data. - No separate sovereign/sub-sovereign PD model due to very low starting PD PiT and small materiality. - Foreign currency (FC) loans and repayment vehicle (RV) loans: - Additional impairment charges calculated for FC and RV loans given susceptibility to home currency depreciation. - Most FC loans in Austria are structured as bullet loans in the form of an RV. - Impairment add-on estimated by segment and country conditional on exchange rate movements, adopting OeNB’s macroprudential stress-test approach. ### Market risk analysis - Market risk captured valuation risks from changes in risk-free interest rates and credit spreads for interest-sensitive instruments, as well as equity and commodity risks. - Regulatory reports at cut-off date provided each bank’s sovereign, corporate and financial issuers’ debt, equity and commodity portfolios by accounting category, maturity bucket, and sovereign issuer. - Macrofinancial scenario extrapolated to produce financial variable shocks (mainly interest rate spread shocks by security type) used to estimate P&L impacts under both scenarios. *Source: IMF staff analysis in the FSAP chapter on Austria (text provided).* ### 34. The analysis covered the impact of interest rate risks and spread risks on sovereign ### 1autea2020006 - 34. The analysis covered the impact of interest rate risks and spread risks on sovereign ### Market risk analysis: scope and methodology - Coverage: impact of interest rate risks and spread risks on sovereign and corporate debt securities in FVTPL and FVOCI accounting portfolios. - Loss treatment: - Losses from FVTPL portfolios assumed to impact regulatory capital through net profits. - Losses from FVOCI portfolios assumed to affect capital through other comprehensive income. - Conservatism: existing hedges assumed ineffective during the scenario horizon. - Exclusions due to data limitations: counterparty credit risk and credit valuation adjustment (CVA) risk were not included. - Valuation approach: - Market valuation losses from interest rate risks derived using a modified duration approach. - Two-step capture: 1. Re-pricing losses in FVOCI and FVTPL books due to shocks to sovereign yield curves. 2. Valuation impact due to shocks to spreads of corporate debt securities. - Spread projections for corporate, bank and financials bonds proxied using average yield per maturity tenor from Bloomberg and anchored to reflect macrofinancial conditions in the two scenarios. ### Other market factors - Equity and commodity price risks: - Equity: all equity positions assumed to pertain to the domestic equity market due to lack of country-breakdown data. - Commodities: applied shocks aligned with the oil shock in the scenario; banking sector exposure to commodities is relatively limited compared to interest rate and spread risks. ### Net Interest Income (NII) analysis: approach and assumptions - Methodology: approach following the spirit of the 2018 EBA Stress Test Methodology, tailored to available data to project NII under the two scenarios. - Asset and liability segmentation (simplified, mapped to COREP and FINREP): - Assets: interbank loans, fixed and variable rate loans, fixed and variable rate bonds, non-interest-bearing assets. - Liabilities: interbank deposits, customer deposits, variable and fixed rate issued bonds, equity. - Data sources to infer starting volumes and effective interest rates: - (i) balance sheet information from national GAAP; - (ii) financial stability reporting of domestically operating entities (OeNB’s integrated reporting framework); - (iii) FINREP reports of foreign subsidiaries complemented by aggregate re-pricing information on fixed rate instruments. - Translation of macro scenario into interest rates: - Scenario translated into shifts in effective interest rates for all segments and countries, including a delta shift on the reference interest rate and a shift in the margin as per EBA approach. - Pass-through and repricing assumptions: - Pass-through constraints on margin components of liabilities and assets set per the EBA stress test methodology. - Maturing instruments assumed to be replaced by identical new instruments (same segment and same initial maturity) but repriced at current reference and margin rates implied by the scenario. - Modeling design: - Semi-structural model using scenario-based effective interest rates and pass-through constraints. - Enables granular NII impact measurement across geographies; requires strong assumptions for effective interest rates of maturing business when granular data are lacking. ### Non-interest income and business risk treatment - Relative weight: total non-interest income during 2018 stood at about 45 percent of net interest income for the entire system. - Stressing of non-interest income: - Constrained approach from ARNIE implementation applied. - Constraints: non-interest income cannot exceed cut-off year (2018) level; additional reductions applied: - Reduction of 15 percent on Net Fees and Commissions Income (NFCI) and Net Trading Income (NTI). - Reduction of 25 percent on dividend income and the share of profit of investments in subsidiaries, joint ventures and associates outside the scope of consolidation. - Administrative and operating expenses floored to their cut-off year levels. - Constraints applied to multi-year, exponentially smoothed averages confirmed by line supervisors. - Business risk charge: - Captures risk from banks with more volatile P&L; component of ARNIE infrastructure. - Calculated based on individual volatility of seven P&L items forming operating profit before risk: NII, NFCI, NTI, Staff Expenses, Other Expenses, Depreciation, Other Profit before Risk. - Cut-off date starting values weighted by historical volatility; not intended to capture low frequency, high impact losses. ### Addressing equity participation risk in decentralized/cooperative sectors - Feature: inverse ownership in Raiffeisen sector (many small local banks together own a few larger banks; multi-tiered ownership). - Contagion channel: lower-tier banks participate in profits and share losses of higher-tier banks; omission would underestimate losses in adverse scenarios. - Modelling assumptions: - Book value of each bank’s equity stake revalued in line with individual participation share in respective upper-tier bank. - Each euro of profit increases book value pro-rata; each euro of loss decreases book value pro-rata. - Losses limited to book value of each equity stake (cannot exceed investment). - Scope: participation risk covered for all banks with participations, but primarily relevant for Raiffeisen sector where no consolidating entity exists; double leverage from participation risk is highlighted as a key vulnerability under stress. - Data: information for participation revaluation sourced from various OeNB business areas and maintained by OeNB’s stress testing team. ### IMF top-down stress test results — main findings - Overall resilience: banks appear resilient to severe macrofinancial shocks; most banks meet hurdle rates at end of stress horizon. - System aggregate CET1 capital ratio: - Baseline: aggregate CET1 capital ratio would increase by 2.3 percentage points by 2021. - Adverse scenario: aggregate CET1 capital ratio would decline by 4.4 percentage points to 11.1 percent by 2021. - Bank-level shortfalls: - Five small banks (out of a total of 440) would have ending CET1 capital ratio lower than minimum CET1 capital requirement in the adverse scenario. - Total capital shortfall under the adverse scenario is approximately EUR 5 million or 0.01 percent of the aggregate system CET1 capital. - Key drivers of capital depletion in adverse scenario: - Credit impairments and lower net interest income are primary drivers. - Credit losses: 4.9 percentage points of CET1 capital under the adverse scenario versus a 1 percentage point decline in the baseline. - Net interest income: on average lower by almost 15 percent for the horizon in the adverse scenario. - Stressed non-interest income and net trading income and a mild difference in market risk losses contribute negatively but are not the main drivers. - Increase in RWA (deteriorating credit quality) contributes an additional CET1 depletion of 0.7 percentage points. - Negative impact partially mitigated by reduced dividend distributions and differences in reduced tax payments because of projected losses. - Heterogeneity across OSIIs and non-OSIIs: - Different drivers of CET1 evolution between large international banks (OSIIs) and smaller domestically-focused banks (non-OSIIs). - Non-OSIIs: equity participation valuation is a key factor—positive contribution of 1.8 percentage points in baseline versus negative contribution of 1.9 percentage points in adverse scenario due to Raiffeisen participation channel. - Participation channel contribution to system capital generation capacity: 0.8 percentage points. - Credit-loss and RWA specifics: - Credit losses: 4.9 percentage points of CET1 capital under adverse scenario; additional 0.6 percentage points attributed to additional impairment for FC loans across all countries in the adverse scenario. - Increase in credit-linked RWAs contributes to CET1 depletion by 0.7 percentage points for the system; impact fully attributed to OSII group (1 percentage point impact). - The OSIIs’ RWA increase corresponds to an increase of total credit-linked RWAs at end of scenario horizon of approximately 7.3 percentage points. - A negligible effect concentrated in O-SIIs attributable to securitization portfolios. - NII dynamics: - Under the adverse scenario, imperfect interest rate pass-through leads to a reduction of overall average NII by 18.9 percent relative to observed net interest income during the year preceding the cut-off date (2018). - Baseline reduction in NII is 4.7 percent due to milder interest rate fluctuations and absence of idiosyncratic shocks to margins on the liability side. *Source: IMF staff summary of the provided chapter content.* ### 51. OSII banks are more severely hit in terms of their aggregate NII reduction, mainly due ### 1autea2020006 - 51. OSII banks are more severely hit in terms of their aggregate NII reduction, mainly due ### Net Interest Income (NII) impact - OSII group NII reduction in the adverse scenario: approximately 20.1 percent on average for the three years. - Non-OSIIs NII reduction in the adverse scenario: 6 percent. - Main driver: sharper reference rate increases in some non-EU countries to which OSII banks are exposed; higher exposures of OSII banks to more volatile reference rates affect interest income more pronouncedly. - System-wide market risk impact: 1.2 EUR billion (or 0.3 percentage of RWAs), with the first year most severely impacted due to the shape of the interest rate shifts. - Market risk portfolios: overall FVPL and FVOCI portfolios are relatively small and of good quality, with some concentrations noted mostly for non-OSIIs in corporate bonds issued by banks and/or financials and slightly above average equity exposures. ### Methodology caveats and calibration - NII impact should be interpreted with caution given the important influence of the methodology on overall results. - A more realistic approach would incorporate country-specific calibration of interest rate shocks and margin overlays to reflect fragilities on the funding side as opposed to bank-specific spreads. - For exposures under STA (standardized approach) a zero-percentage point increase in RWAs is modeled; only part of OSII credit exposures are under the IRB regulatory approach. - Simplified methodology and phased-in gradual introduction of the minimum haircut underpin moderate non-interest income results. ### Non-interest income impact - Non-interest income projections: 10.1 percent lower for the system aggregate in the adverse scenario relative to the baseline (and the 2018 observed income). - Overall impact on RWAs: accounts for only 0.4 percentage points of overall RWAs due to small relative weight of non-interest income in the Austrian banking sector’s overall income generation capacity. - Impact consistent across OSIIs and non-OSIIs. ### Sensitivity analysis — Raiffeisen IPS and contagion - Exploratory test of the resilience of two-layered IPS (regional and federal) in the Raiffeisen sector used FSAP adverse scenario as base case; additional losses of gradually increasing severity introduced to stress the central institution. - Measured inward spillovers using full system simulation with ARNIE; entities with capital shortfalls identified and capital buffers of IPSs measured against projected shortfalls. - Range of support steps considered: current ex-ante funding; future ex-ante funding; contractually binding levels; fully available excess capital above minimum requirements and regulatory buffers. - Findings: confirms centrality and systemic importance of RBI (central institution); negative developments in material geographies may be exacerbated due to inverse ownership structure and elevated sensitivity of overall sector to central institution performance. - Raiffeisen IPS acts as shock absorber for idiosyncratic shocks but may result in inward spillovers in systemic events due to strong equity linkages and reduced income-source diversification and organic capital generation capacity. ### Low interest rate sensitivity - Sensitivity scenario incorporates EU area interest rate evolution up to mid-September 2019: - Assumes decrease of the three months Euribor (10 years Government bond yield) with respect to end-2018 by 10 (80) basis points from 2018YE to mid-September 2019. - After mid-September 2019, interest rates stay constant until end-2021. - For non-EU CESEE currencies, interest rates assumed constant at end-2018 level over the entire three years analysis horizon. - To isolate low interest rate impact, no increase in banks’ credit spreads is assumed. - Under this scenario, system-wide net interest income declines by 5 percent on average until end-2021. - Driver: legal floors on domestic customer deposits prevent interest expense from declining; maturing long-dated fixed rate assets replaced at lower reference rates compress interest income. ### Concentration analysis - Large exposure (LE) simulation (before credit risk mitigations, after exemptions) suggests concentration risks are, on average, contained. - Capital losses would be moderate (2 percent of RWAs) upon default of three largest exposures. - Banks would meet regulatory capital ratios comfortably following defaults of their ten largest counterparties. - Caveats: simulations do not consider credit risk mitigation measures in place and do not consider potential netting arrangements. - Results sensitive to cooperative sector network structure: - Performing test on gross exposures, 5 percent of aggregate capital would be depleted following default of the largest counterparty (reflecting Raiffeisen tiered structure). - On aggregate, system’s CET1 ratio would remain above regulatory minimum even in defaults of ten largest exposures. - Results conservative as they exclude support mechanism provided by Raiffeisen IPS structure. ### Policy recommendations and stress testing function - Inward contagion amplification mechanism requires enhanced monitoring capacity and oversight; authorities should ensure participation risk is adequately measured and micro- and macroprudential buffers properly calibrated to mitigate system-wide risks from equity participations within the Raiffeisen sector. - Strengthen stress testing function: importance of OeNB’s ARNIE infrastructure demonstrated by ability to run granular stress tests for all 440 consolidated Austrian banks. - Authorities should ensure resources and organizational structure are adequate given complexity and resource intensity of stress testing. - Recommended enhancements to stress testing: - fully integrate sensitivity, contagion, and interconnectedness analysis tools in standard macroprudential stress testing toolkit; - further develop framework to capture second round effects, behavioral elements, dynamic balance sheet stress tests, separation of prudential and accounting layers; - increase modelling granularity to capture geographical dispersion and other risk diversity; - integrate a suite of satellite or core models under same infrastructure as feasible alternatives to standard EBA approach for a complete top-down perspective and targeted methodological approach; - enhance capacity to calibrate scenarios comprehensively targeting Austrian banking sector cyclical and structural vulnerabilities complementary to regular EBA exercises. - Data granularity: while Austrian exposures data are good, significant gaps exist on CESEE exposures; achieving desired data coverage for financial stability purposes requires cooperation across authorities and entities and should be strategically prioritized and tactically accelerated. ### Liquidity stress tests and funding profile (overview) - Tests conducted: Basel III LCR-based test over 30 days (by aggregate currency position and by significant currency) and a cash flow-based liquidity test over three months; exploratory quasi-NSFR test also conducted. - System coverage allowed broad set of scenarios and evaluation of additional liquidity protection from contractual liquidity support commitments of two-layered IPS in decentralized Raiffeisen sector. - Funding mix: retail and corporate deposits account for almost 40 percent of banking system's liabilities. - Repo segment: does not play a significant role. - Derivatives-related liabilities: account for about 18 percent of total funding, importance due to Austrian banks’ presence in CESEE region. - Asset and liability derivative positions: mostly linked FX and cross-currency swaps and seem to be balanced. - Regulatory reporting provided granular data including contractual maturities, asset counterbalancing capacity, and collateral encumbrance facilitating liquidity monitoring and analysis. *Source: IMF staff and OeNB material as presented in the provided content.* ### 70. The LCR metric measures the ability of banks to meet liquidity needs in a 30-day ### 1autea2020006 - 70. The LCR metric measures the ability of banks to meet liquidity needs in a 30-day ### LCR-based stress tests: design and scenarios - The LCR metric assesses ability to meet liquidity needs in a 30-day liquidity stress scenario using a stock of unencumbered high-quality liquid assets (HQLA). - Regulatory standard: 100 percent. - FSAP calibrated three scenarios measuring banks’ ability to withstand a 30-day stressed run-off rate against the regulator-prescribed run-off rate for the standard LCR calculation: - Scenario 1: Stressed run-offs on retail deposits (elevated retail withdrawals). - Scenario 2: Increased run-offs on wholesale sources of funding, especially wholesale and corporate demand deposits. - Scenario 3: Combination of Scenario 1 and Scenario 2 plus an additional decline in the price of liquid assets (market price shock). This scenario is described as less likely because stress factors are applied in parallel for most funding segments across all banks. - Annex VI contains a summary of assumptions and deviations from regulatory LCR parameters (as referenced). ### LCR-based stress tests: key results - Asset-weighted average LCR: 164.7 percent. - Outflows-weighted average LCR: 148 percent. - Asset-weighted LCR under retail funding outflows scenario: 128 percent. - Asset-weighted LCR under wholesale funding outflows scenario: 119.5 percent. - Asset-weighted LCR under the third (“severe”) scenario (including market price shock): 104.3 percent. - OSII banks have smaller liquidity buffers than non-systemic institutions; average OSII LCR falls below 100 percent only in the third scenario, while the weighted-average OSII LCR remains very close to the regulatory minimum even in that scenario. ### Currency-specific LCR results - EUR LCR: declines from 145.8 percent under the regulatory definition to 96.9 percent under the severe scenario. - U.S. dollar (USD) LCR: - System-wide USD LCR average at the starting level: 56.6 percent (below the 100 percent monitoring mark). - USD LCR in the severe scenario: 44.1 percent. - USD outflows in the severe scenario: amount to 6.1 billion. - Interpretation: USD is not a significant currency for the Austrian banking system; USD outflows remain manageable from a systemic perspective. ### Cash-flow analysis: design - Horizon: 3 months of stressed liquidity conditions. - Based on contractual maturity ladder for assets and liabilities; highest levels of roll-over needs by liabilities segment during a forward-looking 3-month period were identified and applied. - Scenarios vary by run-off rates on liabilities and fair value and discount haircuts on liquid assets; increasing severity applies higher run-offs and haircuts. - Partial utilization of off-balance-sheet credit or liquidity commitments is accounted for. - Counterbalancing capacity measured by additional liquidity generation across liquid asset classes after scenario-specific haircuts and compared to funding needs to establish liquidity surplus/shortfall. - Full collateral revaluation is performed for each scenario, including collateral used for repo and reverse repo transactions. - Cash-flow test assumption: banks continue to provide credit to the real economy; roll-over rates of maturing retail and corporate loans are assumed to be 100 percent (i.e., banks are not allowed to reduce new credit extension to counterbalance outflows). This increases severity to ensure balance sheet capacity to continue lending under stress. - Cash-flow test also accounts for second-order effects from reduced asset valuations (e.g., margin calls for collateralized funding positions). ### Cash-flow analysis: key results - Cash-flow tests identify small liquidity shortfalls for some small banks under the severe scenario. - Combined liquidity shortfall in the extreme (severe) case with IPS support: EUR 1.5 billion. - Aggregate net liquidity position (counterbalancing capacity minus net outflows) remains positive across scenarios: - Initial level: 21 percent of system assets. - Severe scenario: 8.8 percent of system assets. - Drivers of resilience: elevated level of high-quality securities holdings and stable funding mix dominated by customer deposits. - OeNB’s ability to provide liquidity through standard facilities or extraordinary measures renders the EUR 1.5 billion shortfall manageable. ### IPS (Internal Payment/Protection System) scheme analysis - Cash-flow test repeated under scenario assuming no additional IPS liquidity protection (Raiffeisen banks withdraw contributions and contractual obligations terminated). - Under no-IPS scenario, combined liquidity shortfall increases to EUR 1.7 billion (compared to EUR 1.5 billion with IPS support). - Conclusion: IPS enhances resilience to liquidity shocks, but the banking system can withstand prolonged adverse liquidity conditions even without IPS support. Some regional IPSs show shortfalls in the severe scenario, but federal IPS layer buffers are adequate to support regional shortfalls. ### NSFR exploratory stress test - Proxy NSFR computed based on NSFR monitoring templates and EBA weights. - Simple stress: funding market remains dysfunctional for a year; ASF liabilities maturing during next four quarters assumed rolled-over as short-term liabilities; some migration of ASF balances with maturities above one year into Q1–Q4 maturity buckets to mimic historical volumes. - Stressed NSFR uses original RSF volumes and proxy weights but stressed ASF volumes. - Asset-weighted aggregate NSFR stands at 122 percent. - System-wide NSFR under stress: 112 percent. - Share of system assets with NSFR below 100 percent: - Initial: 2.2 percent of system assets. - Under stress: increases to over 28 percent of system assets. - Timing note: NSFR will come into force in 2021Q2. ### Policy recommendations (liquidity toolkit enhancements) - Further develop systemic liquidity components and enhance the macroprudential angle: - Develop approaches or calibrate parameters that deviate from the microprudential angle where systemic assumptions on run-offs and haircuts may differ. - Enhance forward-looking aspects of cash-flow analysis by assessing cliff effects over time bands: - Using the last reporting date as a snapshot may create blind spots that impair resilience assessment. - Enhance granularity of asset-side data regarding encumbrance, eligibility for central bank liquidity facilities, and fair value quality characteristics (duration, issuer credit quality, present haircuts, etc.): - Facilitates development of a layer for second-round effects (fire sales, realistic modelling of market haircuts, simulation of LoLR policy decisions). - Fully integrate analysis of liquidity support buffers due to IPS structures into standard liquidity analysis by defining and quantifying IPS impact on systemic liquidity for distinct support scenarios for failing IPS members. ### Contagion analysis: overview and methodology - Austrian banking system characterized as heavily tiered and clustered with a few important central nodes and many smaller banks; focal nodes include central institutions of liquidity associations in decentralized sectors. - Network analysis module assesses direct contagion from individual hypothetical defaults: - A bank’s induced failure leads to default on all interbank obligations. - Failing bank can cause liquidity strains to borrowing counterparties through forced fire sales, increasing potential for indirect defaults. - Subsequent defaults occur when a bank breaches minimum capital requirement (total capital adequacy ratio drops below 4.5 percent of total risk-weighted assets). - Two channels of distress modeled: - Cascade effects: default of one bank triggers assessment of defaults of other banks iteratively. - Funding-replacement fire-sale losses: affected bank may sell assets at a discount to replace lost funding, causing additional losses. - Caveat: analysis does not consider interbank netting arrangements. ### Contagion analysis: data and calibration - Analysis conducted on consolidated gross exposures from central credit registry covering credit instruments above EUR 350,000 (loans, guarantees, commitments; excludes equity). - Exposures to affiliated banks within cooperative sector were deducted for exercise purposes due to liquidity reserve structures. - Assumed parameters (as presented in the content): - λ1=0.550 percent LGD on unsecured lending - λ2=0.2 20 percent LGD on secured lending - ρ=0.35 Share of lost funding that is non-replaceable - δ=0.3 30 percent discount on asset sales - capital CET1 capital under Basel III - bank default CET1 capital falls below 4.5 percent - Assumed additional values described in text: - 65 percent roll-over ratio of interbank liabilities - 80 percent recovery rate for secured interbank lending - 50 percent recovery rate for unsecured lending - 30 percent discount in fire sale of assets (δ=0.3) ### Contagion analysis: key results and sensitivity - Results sensitive to treatment of cooperative sector exposures: - When network uses gross bilateral exposures including affiliated entities (i.e., cooperative sector exposures included), contagion and vulnerability indices are considerably higher. - Maximum contagion index for group of OSII banks on gross exposures: 41 percent of total capital (unweighted). - Median bank contagion for OSII (gross exposures): 1.1 percent. - Excluding exposures across affiliated entities: - Maximum contagion falls to 1.2 percent of system capital. - Median contagion decreases to 0.5 percent. - Effects are highly skewed, highlighting the central institution’s key role in the decentralized Raiffeisen sector. - Caveat: equity exposures are not included and would amplify interbank losses if present. *Source: IMF staff calculations and analysis as presented in the provided content unit.* ### 86. The FSAP team examined interbank cross-border contagion of Austrian banks using ### The FSAP team examined interbank cross-border contagion of Austrian banks using the CoMap methodology. ### Analysis and methodology - The analysis studied the impact from counterparty default risk and liquidity pressures from linkages to domestic and international banks. - Bank specific calibrated parameters were used to measure banks’ systemic importance and their degree of fragility (for example, banks’ funding rollover needs and liquidity buffers). - The CoMap approach was applied at the consolidated level and hence excludes intragroup contagion. - The index of contagion shows the percent of total capital impairment in the banking system due to the failure of each bank. The index of vulnerability is the average percent capital impairment for a bank due to the failure of other banks. ### Key findings on contagion and spillovers - Results suggest that outward and inward spillovers between Austrian banks’ and foreign banks in the CESEE and EA are limited. - Limited spillovers are partly driven by exemptions in the LE regulatory reporting template which eliminate most exposures among affiliated parties in the decentralized sectors, as well as cross-border intragroup exposures. - Liquidity buffers held by Austrian banks limit the extent of contagion from the default of a funding counterparty in Austria. - Median and percentile contagion statistics: - Median contagion index from Austrian OSII banks to the Austrian banking system: 0.83 percent. - Contagion at the 10th percentile: 0.5 percent. - Contagion at the 90th percentile: 2.1 percent. - The non-OSII sample shows negligible contagion effects. - Cross-border interbank contagion effects: - Outward spillovers to the CESEE are limited to a few Austrian banks. - Median contagion from Austrian OSIIs to the EA: 0.12 percent capital depletion. - Inward spillovers to Austrian banks from hypothetical outright bank defaults in the CESEE and Italy are negligible. ### Policy recommendations on contagion analysis - Conduct network analysis on a detailed breakdown of exposures by instrument to obtain a more precise picture of contagion and vulnerability. - Move beyond the two extreme scenarios used (one in which all exposures within the Raiffeisen sector are subject to default, and one in which no exposures are lost within the Raiffeisen network) toward an exercise closer to reality: - Exclude only liquidity reserves from counterparty credit risk (i.e., those held at the central institution). - Include equity exposures in the exercise. - Monitor the value of augmented “net” exposures closely over time, possibly on a quarterly basis, and feed these into network analysis to monitor changes in contagion and vulnerability over time. ### Stress testing: scope, channels, and scenarios - Institutional perimeter and data: - Institutions included: All Austrian credit institutions both directly supervised by the ECB (six SIs) plus one subsidiary of a foreign SI) and directly supervised by the Austrian authorities (433 LSIs). - For the 7 SIs, about 60 percent of banking sector assets. - For the entire 440 institutions, above 95 percent of total assets of deposit-taking institutions in Austria. - Data as of December 2018. - Data sources: European and Austrian regulatory returns and supervisory data (e.g. FINREP and COREP); OeNB’s statistical data warehouse; Austrian Central Credit Register (CCR); external data sources (Kreditschutzverband, KSV); Moody’s Analytics: CreditEdge data on corporate default probabilities. - Scope of financial consolidation: group-wide. - Channels of risk propagation and modelling: - Balance sheet approach; projections of key balance sheet, income statement and capital account items; static balance sheet assumption. - The exercise will not provide separate accounting projection layers for impairments. Only 19 entities from the sample report under IFRS 9 with the remaining reporting under national GAAP. - Credit risk, market risk, net interest income and non-interest income projections for two scenarios: baseline and macro adverse. - Indirect credit risk from foreign currency loans and repayment vehicles is stressed for CHF, JPY, USD and EUR. - Participation risk accounted for based on OeNB’s internal entity equity participation matrix. - Granular projections of credit risk parameters including EADs, PDs, and LGDs for each asset class and geography. - Satellite models and other modelling techniques: - Different asset class segmentation for SIs and LSIs; LSI segments mapped to a CCR relevant segmentation including granular corporate exposures by NACE classification. - Net interest income projected based on sensitivity to macrofinancial conditions and repricing profiles; approach similar to EBA exercise. - Net trading income, net fee and commission income stressed based on historical volatility and EBA-based haircuts. - Operational expenses kept at the starting point level. - Mark-to-market approach used for equity and commodity price impacts on net open positions; debt securities at Amortized Cost portfolios are not stressed. - Structural model approach and Monte Carlo simulations used to estimate PDs and LGDs for mortgage exposures across geographies. - Bayesian Model Averaging (BMA) techniques used to produce satellite projections for the corporate sector across geographies. - Tail shocks and scenarios: - Stress test horizon: Three years (2018 Q4 – 2021 Q4). - Two scenarios: baseline and macro adverse. - Baseline scenario based on July 2019 World Economic Outlook (WEO) projections. - Macro adverse scenario calibrated using the Global Macrofinancial Model (GFM) and assumes materialization of systemic risks highlighted in the RAM, including contagion from CESEE countries, a sudden sharp tightening in global financial conditions, a correction in real estate prices in Austria, and a slump in global growth. - Macro adverse scenario severity: implies a deviation of Austria real GDP from its baseline of 6.9 percent by 2021, with a 2.3 Standard Deviation move in two-year cumulative real GDP growth rate, and a 20 percent peak-to-trough decline in real estate prices. Output shocks in the CESEE region range between 8.1 and 12.6 percent deviation from baseline. - Sensitivity analysis: Low-for-Long interest rate scenario used as adverse interest scenario; reverse stress tests performed; complementary simulation analysis of the Raiffeisen bank inverse ownership structure to measure participation risk and inward spillovers. ### Behavioural assumptions, risks, and buffers in the stress test - Credit risk captures all on-balance/off-balance sheet exposures at amortized cost by regulatory exposure sector and geography; starting points of credit parameters used to project scenario dependent forward paths. - Market risk reflected in valuation effects of FVTPL and FVOCI positions, and net open positions; interest rate curves by country inferred from scenario rates. - Net interest income affected by projecting effective interest rates by asset/liability class and geography; a time to repricing approach based on fixed/variable rate breakdown informs pass-through velocity. - Shocks to non-interest income simulated to capture market-sensitive components. - Risk weighted assets adjusted to reflect changes in credit exposure quality. - Behavioral adjustments and constraints: - Static balance sheet assumption: exposures remain constant and do not evolve with scenario credit growth assumptions. - For NII, maturing assets/liabilities are assumed to be replaced by instruments of the same type and maturity but at current rates. - If banks’ capital falls below regulatory requirements, no prompt corrective action is assumed. - Banks are assumed to pay a fixed share of 30% of their profits, if positive, in taxes and another 30% as dividends to shareholders. - One-off adjustments made by OeNB line supervisors as part of the OeNB’s annual stress testing exercise will be accepted to warrant a uniform starting point at T0. *Source: IMF staff calculations and FSAP team analysis as presented in the provided content.* ### 5. Regulatory and ### 5. Regulatory and market-based standards and parameters ### Calibration of risk parameters - Scenario dependent forward paths for PiT PDs and LGDs are estimated for each asset class and geography. - It is assumed that prudential expected losses will coincide with accounting impairments; therefore, the accounting layer is ignored. - Given the limited availability of PiT LGD data, some very basic proxies are used. - For internal ratings-based (IRB) exposures, risk-weight assets are projected based on updated regulatory TTC PDs and downturn LGDs, using appropriate scaling multipliers from the PiT parameters. - For standardized approach (STA) exposures, risk-weight assets are assumed to remain constant. - In the baseline, hurdle rates include the regulatory minimum and all applicable buffers. - In the adverse scenario, the regulatory minimum is assumed to be the hurdle rate, i.e. all buffers can be drawn-down. - Hurdle rates are based on the common equity tier-1 ratios. ### Reporting format for results - System-wide evolution of CET1 capital ratios. - Distribution of banks’ capital positions. - Contribution to key drivers to system-wide net income and capital position, including differences between the baseline scenario and the adverse scenario. - Share of institutions with capital below the hurdle rates. ### Infrastructure used - OeNB’s ARNIE infrastructure is used to account for the extended bank sample and the cross-entity equity participations (inverse ownership). - IMF team’s satellite model projections are imported as an external overlay into ARNIE. - Banks’ credit and interest rate starting point parameters were validated using IMF staff estimates and in some cases a scaling factor was applied to the satellite model to anchor projections better with regulatory or historically observed parameters. - Starting point translation into scenario dependent forward paths for individual banks and segments follows the IMF guidelines (absolute shift, distance to defaults or similar type of translation). --- ### Bank Liquidity Stress Testing ### 1. Institutional perimeter - Institutions included: Seven SIs, and 433 Austrian LSIs. - Market share: - For seven SIs, about 60 percent of banking sector assets. - For all 440 entities (including the 7 SIs), above 95 percent of total assets of deposit-taking institutions. - Data and baseline date: - ECB/SSM and OeNB: regulatory returns based on the Liquidity Coverage Ratio and the Net Stable Funding Ratio and Additional Liquidity Metrics from the FINREP/COREP data repository. - Data as of December 2018. - Scope of financial consolidation: group-wide. ### 2. Channels of risk propagation - Methodology: - The exercise is based on three types of tests—LCR test, cash-flow analysis and NSFR test. - The LCR test is in line with the standard Basel monitoring tool, featuring total liquidity and liquidity in all significant currencies (Euro, Swiss Franc, US dollar and CESEE currencies). - The cash-flow analysis analyzes the net cash balance, accounting for available unencumbered assets, contractual cash inflows and outflows, and behavioral flows. - For the cash-flow analysis, relevant second-round effects could be considered, including margin calls for existing collateral positions, central bank’s liquidity provision, additional asset haircuts due to fire sales, additional repo haircuts due to limited collateral supply, and wholesale funding market freezes because of banks’ solvency and liquidity concerns. - NSFR reporting is still used for monitoring purposes (non-binding). The analysis reports recent NSFR statistics but also introduces standard parameters for the calculation of stressed NSFR. - The IPS structure was taken into account -to the extent possible- when assessing liquidity under stress. - Regulatory liquidity waivers were considered, and the mandatory interbank deposits of Article 27(a) of the Banking Act were reallocated to the depositing entities, increasing their counterbalancing capacity. - Satellite models for macrofinancial linkages: - For the cash-flow analysis, asset haircuts reflect two components: (i) shocks to interest rates and asset prices as captured the macrofinancial scenarios; and (ii) additional haircuts required by counterparties to accept specific assets as collateral for secured funding transactions. - Stress test horizon: - For the LCR test, the stress test horizon is 30 days. - For the cash-flow analysis, the horizon of stress events would normally be 3 months. Nonetheless, a longer period of stress events (up to 1 year) may be considered as sensitivity analysis. ### 3. Tail shocks - Scenario analysis: - For the LCR test, twelve scenarios are considered as a combination of: (i) three scenarios on liquid assets shock (regulatory, mild and severe), and (ii) four scenarios on liability outflows; regulatory, one reflecting retail outflows, one reflecting higher wholesale outflows, and one combining the retail and wholesale outflows. - For the cash-flow analysis, a series of scenarios are considered, with a range from mild to severely adverse liquidity conditions. The cash-flow analysis considers both funding and market liquidity risks. - For the NSFR analysis, only one stress scenario featuring inability of banks to roll=-over longer-term funding positions for a period of a year. (still of exploratory nature, given the smaller experience on stressed NSFR). - Sensitivity analysis: N/A ### 4. Risks and buffers - Risks/factors assessed: - Funding liquidity risk is reflected in funding run-off rates and asset roll-over rates, the latter providing cash inflows related to non-renewal of maturing assets. - Market liquidity risk is reflected in asset haircuts, which could be influenced by market movements, potential fire sales and collateral supply considerations. - Behavioral adjustments: - Liquidity from the central bank’s emergency lending assistance (ELA) is not considered. - The cash-flow analysis may consider some behavioral assumptions about a counterparty’s ability or willingness to transact based on banks’ solvency and liquidity conditions. ### 5. Regulatory and market-based standards and parameters - Calibration of risk parameters: - The LCR tests are based on regulatory and stress parameters. - The cash-flow analysis may incorporate relevant second-round effects. - Stress funding run-off rates, asset roll-over rates, and asset haircuts are calibrated based on empirical evidence and relevant international experiences. - Regulatory/accounting and market-based standards: - LCR per Basel III; the hurdle at 100 percent (at the aggregate currency level). - Net cash balance for the cash-flow analysis; to pass, a non-negative net cash balance is required, where the balance reflects net funding outflows and counterbalancing capacity. - NSFR is not yet applicable but a targeted post-introduction limit of 100 percent was assumed. ### 6. Reporting format for results - Changes in the system-wide liquidity position, including important drivers for cash outflows, cash inflows and counterbalancing capacity. - Distribution of banks’ liquidity positions. - Number of institutions with LCR/NSFR below 100 percent and/or negative net cash balance. - Amount of liquidity shortfalls, including by currencies. ### 7. Infrastructure - Fully comprehensive infrastructure developed my IMF staff with a FINREP/COREP data repository integrated backbone. --- ### Network Analysis ### 1. Institutional perimeter - Institutions included: Seven SIs, and 433 Austrian LSIs. - Market share: - For seven SIs, about 60 percent of banking sector assets. - For all 440 entities (including the 7 SIs), above 95 percent of total assets of deposit-taking institutions. - Data and baseline date: - Austrian central credit registry data. - Data as of December 2018. - Scope of financial consolidation: group-wide w.r.t. to Austrian subsidiaries. ### 2. Channels of risk propagation - Methodology: - Network analysis using Furfine algorithm and Espinosa-Sole tool. - Includes contagion channels from funding concentration and foreign counterparties. - Cascading effects from individual defaults through credit and funding counterparties. - Linkages with solvency and liquidity stress tests: - The transmission of funding shocks is linked to liquidity stress test results by allowing banks to draw down their liquid buffers to replace funding from defaulting funding counterparties. - Buffers: - Tier 1 capital. - Counterbalancing capacity. ### 3. Tail shocks - Size of the shock: Outright defaults. ### 4. Sensitivity test factors - Performance of collateral (for secured exposures). - Loss given default (for unsecured exposures). - Role of netting arrangements. - Elimination of exposures within the decentralized Raiffeisen sector. - Use of CoMap tool to account for Austrian banks’ idiosyncratic calibrations and spillover risks to/from foreign counterparties using the large exposure database, gross exposure net of exemptions. ### 5. Reporting format for results - Failed capital in percent of total capital. - Contagion index. - Vulnerability index. - Grouping of banks by OSII (for contagion index), and SyRB (for vulnerability index). --- ### Annex III. Credit Risk Satellite Models — Probabilities of Default Modelling (PD PiTs) ### Overview and approach - A series of econometric models were estimated to produce scenario-dependent forward paths for point-in-time (PiT) probabilities of default (PDs) for the different exposure segments. - The selection of the modelling approach was largely driven by the loans and advances segmentation and the availability of corresponding historical time series for calibration purposes. - Given the very limited availability of exposure segment specific historical data and the need to ensure robustness, the calibration of PD models for the corporate and financials segments were based on historical data from a third-party provider. - The lack of reliable historical data on mortgage and unsecured retail credit exposures was the reason for deploying a structural model for the calculation of mortgage loss rates across all material geographies. The methodological approach and the resulting models are described in Annex IV. - For the solvency stress test, the FSAP team estimated satellite PD models for a total of 18 material geographies, which accounted to approximately a model coverage exceeding 90 percent of the exposures. ### Corporate and Financials segments - A Bayesian Model Averaging (BMA) econometric technique was employed for modelling and projecting the default rate at the individual geography and portfolio segment levels. - The BMA approach operates with a pool of equations (several hundreds or thousands) used to estimate the default rates. Weights are assigned to each equation that reflect their predictive power, resulting in a “posterior model” equation. - The pool of equations contains equations for every single credit risk indicator (per portfolio segment and geography), by considering all possible combinations of predictors from a pool of potential predictor variables, including variables such as real GDP, investment, consumption, exports, price inflation, and short- and long-term interest rates. - Table III.1: Geographies for which a model is estimated (18 geographies listed): 1 AT Austria 2 CZ Czech Republic 3 SK Slovakia 4 HU Hungary 5 RO Romania 6 BG Bulgaria 7 SI Slovenia 8 HR Croatia 9 PL Poland 10 RU Russia 11 UA Ukraine 12 DE Germany 13 UK UK 14 ES Spain 15 FR France 16 US US 17 TR Turkey 18 ROW Rest of the World ### Modelling details - To ensure PD predictions between 0 and 1 and to capture nonlinearities, a logit transformation was applied to the original PD. - The logit-transformed PDs were modeled as a linear function of exogenous macroeconomic and financial factors (regressors). - The model specification allows inclusion of autoregressive lags as well as lags of the explanatory variables, to account for the backward-looking nature of credit risks. - The estimated conditional PD PiT forecasts for each segment estimated by the BMA models were further anchored to the aggregate (system-wide) starting point PDs to eliminate any potential overshooting attributable to the market-based nature of the historical default rate time series. - Bank-specific conditional forward paths were then produced using a translation of the system-wide forward path model using an absolute shift in the PD space. ### Inputs, calibration and selection criteria - Explanatory variables used include real GDP, exchange rate, inflation, unemployment rate, output gap, stock market price indices as well as outstanding private credit. - Inputs other than unemployment rates are subject to quarterly growth transformation, and unemployment rate were taken as percentage point changes between periods. - The sample period used for calibration ranges between 2000 and 2018, and a quarterly frequency was used. - Model selection for the BMA follows several criteria: - R-square. - Durbin Watson statistics (target between 1.5 to 2.5). - Number of significant variables with high posterior inclusion probability. - Quality of in-sample forecast. - Size of the impact in the forecasting period. - The ideal candidate would have a relatively high R-square, a Durbin Watson statistic between 1.5 to 2.5, a small root-mean-square-error, and a historically consistent size of impact under stress. ### Key findings from PD models - The scenario dependent PD projections are broadly in line with historical stress episodes for the corporate and the financial segment, with the size of impact broadly in line with past stress episodes (in particular during the GFC). - Results display salient idiosyncrasies among segments, with corporate segment generally displaying higher PDs, for both in-sample realizations and out-of-sample forecast, across all countries. - Output, unemployment rate, equity returns, and credit-related measures play significant roles in the determination of underlying credit risks, reflected in high posterior inclusion probability and sizable long run multiplier estimates for both segments and most geographies. *Source: 1autea2020006 - 5. Regulatory and market-based standards and parameters* ### 12. Separate models were estimated for all material geographies of the all material ### 12. Separate models were estimated for all material geographies of the all material geographies ### Model estimation approach and data proxies - Separate PD PiT (point-in-time) models were estimated for 13 material geographies for corporates and financials. - Satellite proxies used where calibration or historical series were limited or unavailable: - SK and SI proxied using the CZ satellite forward path. - BG and HR used the RO path. - ROW mapped to a time series corresponding to EA. - For CZ a single corporate-segment historical series was available; the same forward path was applied for corporates and financials. - Model calibration cross-checked using: regulatory PDs and historical evidence from observed defaults during past downturns. - PD path trajectories under baseline and adverse scenarios illustrated in Figure III.1 (corporates) and Figure III.2 (financials). - Table III.2 reports equation estimates and key statistics for all models estimated using the BMA methodology. ### Key PD PiT model estimation results (Table III.2 — selected coefficients and statistics) - Dependent variable: Moody’s EDF. Results shown as normalized long-run multipliers. Notes: “*” denotes higher posterior inclusion probability than prior (variable statistical significance). - Austria (Corporate / Financials / All): - GDP growth, yoy: -0.83* -0.41* -0.78* - Stock price index, yoy: -0.42* -0.48* -0.54* - R square: 0.66 0.7 0.7 - Number of observations: 48 48 48 - Czech Republic (Corporate / All; Financials group columns blank for some entries): - Private credit, yoy: 0.91* 0.87* - R square: 0.83 0.82 - Number of observations: 64 64 - Hungary: - Inflation, yoy: -0.33* -0.01 -0.27* - Output gap, percent: -0.65* 0 -0.55* - Unemployment rate, yoy (Financials): 0.29* - R square: 0.71 0.09 0.52 - Number of observations: 67 67 67 - Romania: - GDP growth, yoy: -0.55* -0.43* -0.6* - Unemployment rate, yoy: 0.05* 0.27* 0.03* - Private credit, yoy: -0.27* -0.16* -0.25* - R square: 0.89 0.83 0.88 - Number of observations: 36 36 36 - Poland: - Stock price index, yoy: -1.19* -0.91* -1.25* - R square: 0.92 0.83 0.93 - Number of observations: 60 60 60 - Russia: - Inflation, yoy (Financials): -0.19* - Output gap, percent (Financials): -0.56* - Stock price index, yoy: -0.36* -0.2* -0.36* - Private credit, yoy: -0.64* -0.03 -0.64* - R square: 0.76 0.64 0.76 - Number of observations: 66 47 66 - Germany: - Stock price index, yoy: -0.67* -0.19* -0.65* - Private credit, yoy: -0.48* -0.58* -0.49* - R square: 0.78 0.59 0.76 - Number of observations: 67 67 67 - Euro Area: - Stock price index, yoy: -0.82* -0.24 -0.84* - Private credit, yoy: -0.4 -0.34* -0.42 - R square: 0.93 0.8 0.92 - Number of observations: 67 67 67 - United Kingdom: - Unemployment rate, yoy: 0.28* 0.64* 0.41* - Stock price index, yoy: -0.67* -0.2 -0.52* - R square: 0.88 0.86 0.89 - Number of observations: 67 67 67 - Spain: - Output gap, percent: -0.39* -0.02 -0.31* - Stock price index, yoy: -0.42* -0.22* -0.38* - Private credit, yoy: -0.43* -0.61* -0.52* - R square: 0.91 0.9 0.94 - Number of observations: 67 67 67 - France: - GDP growth, yoy (Financials): -0.66* - Exchange rate against USD, yoy (Financials): 0.28* - Stock price index, yoy: -0.78* -0.31* -0.77* - R square: 0.58 0.65 0.59 - Number of observations: 67 67 67 - United States: - GDP growth, yoy (All): -0.06* - Unemployment rate, yoy (Financials): 0.5* - Output gap, percent (Financials): -0.51* - Stock price index, yoy: -0.44* -0.04 -0.64* - Private credit, yoy: -0.44* -0.02 -0.32* - R square: 0.63 0.77 0.74 - Number of observations: 67 67 67 - Turkey: - Unemployment rate, yoy: 0.4* 0.35* 0.38* - R square: 0.17 0.16 0.16 - Number of observations: 67 67 67 Notes on Table III.2: - Number of lags of independent variables is 2 for all reported equations. - Long-run multipliers are normalized: defined as sum of contemporaneous and lagged coefficients and multiplied by ratio of standard deviations (left-hand-side/right-hand-side). Default rates included in logit format. - Asterisks denote variables with higher posterior inclusion probability than prior (statistical significance). ### Loss Given Default (LGD PiTs) modelling - LGD PiT modelling follows OeNB’s internal approach for stressing LGDs. - Effective LGD for each asset segment = LGD_unsec × (1 - collateral ratio), where collateral ratio = collateral / total exposure; collateral assumed LGD = 0. - Only collateral eligible under Basel credit risk mitigation regime is recognized. - Collateral is shocked by applying haircuts based on collateral type and relevant scenario shock, increasing effective LGD under stress. - LGD_unsec calibration: - For Austrian exposures: calibrated using central credit register data. - For other countries: based on World Bank’s Doing Business database, using the Austrian LGD as an anchor. - In the adverse scenario LGD_unsec is shocked using scaling factors anchored to the applied macro scenario. ### Structural model for mortgage defaults (Annex IV) — design, assumptions, and implementation - Rationale: - Structural model used as a satellite due to lack of reliable historical default time series for mortgages across all countries and gaps in borrower-based metrics. - Austrian credit register: all exposures below EUR 350k captured as pooled bucket “KK” with no distinction between mortgages and other retail credit. - Key assumptions made for deploying the structural model: - A DSR/LTV joint distribution for the existing mortgage stock was constructed assuming an average maturity of individual loans of 25 years at origination and an unchanged distribution for the years prior to 2011. - For vintage years prior to 2011 the 2011 vintage DSR/LTV distribution was assumed to be the actual one. - DSR and LTV ratios across vintages adjusted using household income and house price historical time series as proxies for impact on DSR and LTV respectively. - Existing liquid wealth buffers were not used to avoid default under stress in the construction of distributions. - The same joint DSR/LTV distribution was assumed for all material geographies; geographic differences arise from country-specific scenario shocks (interest rate, unemployment, house prices). - Model mechanics: - Uses TUI model (Reserve Bank of New Zealand) with adaptations; has been applied in several FSAP missions and for macroprudential calibration. - Structural probability of household distress, P_distress_t, is a function of: - Demographic distress component (DD), - Borrower’s post-stress DSR (DSPR_t) and change in DSR (ΔDSPR_t), - Unemployment rate u_t and change in unemployment Δu_t, - Equation form provided as (Eq.1) in the source (structural functional form with parameters a0, a1, β1, α1, β2, α3, α4, α5, β3). - Default occurs only if: - Household is in distress (first stage), - Household’s liquid wealth is not enough to cover servicing needs, - Post-stress LTV > 1 (negative equity prerequisite). - Default probability formula (Eq.2) combines: - Probability that property value after stress minus liquidation discount C is lower than outstanding loan notional L, - Probability that the behavioral survival rule B_t(x) (use of liquid wealth LWL_0 at cut-off) fails to prevent default. - Behavioral rule calibration used in solvency stress test: linear survival rule between wealth buffer of 8 and 36 months (buffers below 8 months insufficient; buffers above 36 months yield survival probability 1). - Model outputs and interpretation: - The model accounts for household affordability evolution, house price shocks, and borrower behavioral assumptions to estimate scenario-dependent default and loss rates. - Monte Carlo simulation framework allows incorporation of multiple behavioral rules and house price shock distributions. - Under the described scenarios and initial distributions, the documentation notes “moderate losses for both scenarios for most countries” (textual observation tied to model runs and Figure IV.1). *Source: IMF staff calculations.* ### 6. The conditional LGD is driven by the discounted sale price of the house. The sale occurs ### 6. The conditional LGD is driven by the discounted sale price of the house. The sale occurs ### Conditional LGD formula and mechanics - Conditional LGD at time t is given by (Eq.3): - 퐿퐿퐿퐿퐷퐷 푡푡 = 1 − (1 − 훿훿) * 푃푃 푡푡+푠푠 / [퐿퐿∗(1+푖푖 푡푡 + 푐푐푖푖 푡푡 ) 푠푠] - Definitions and assumptions embedded in the formula: - Sale occurs at time t+s, where s denotes the average time to realize the collateral. - Sale proceeds are net of transaction costs (discounted at a rate reflecting the scenario interest rate premium). - An additional foreclosure liquidation discount 훿훿 is applied. ### Partitioning, simulation approach, and mapping to portfolio loss rates - Joint distribution partitioned along DSR and LTV dimensions. - LTV partitioning values: 1.2, 1.1, 1.0, 0.9, 08, 0.7, 0.6, 0.5. - DSR partitioning values: 0.2, 0.3, 0.4, 0.5, 0.6. - The first two dimensions of the household data are mapped to a 5 by 8 partitioned space. - For each DSR/LTV partition a Monte Carlo simulation (on house price changes anchored to a central house price shock) is used to produce model-based projections on 3-year loss rates. - A portfolio average 3-year loss rate corresponds to the weighted average of the projected loss rates per DSR/LTV density partition. ### Table IV.1 — Loss Rates Estimated for Material Geographies (Pre-Starting Point Adjustments) - Loss Rates (푷푷푷푷∗푳푳푳푳푷푷) (in percentage points) - Baseline 3 Year totals (2019, 2020, 2021, Total 3 Year): - Austria: 0.13, 0.13, 0.13, 0.39 - Czech Republic: 0.31, 0.22, 0.39, 0.92 - Slovakia: 0.03, 0.03, 0.03, 0.08 - Hungary: 0.10, 0.10, 0.10, 0.30 - Romania: 0.03, 0.03, 0.03, 0.08 - Bulgaria: 0.06, 0.06, 0.06, 0.19 - Slovenia: 0.02, 0.02, 0.02, 0.05 - Croatia: 0.11, 0.11, 0.11, 0.34 - Poland: 0.05, 0.05, 0.05, 0.16 - Russia: 0.10, 0.08, 0.11, 0.29 - Ukraine: 0.01, (0.00), 0.01, 0.02 - Germany: 0.03, 0.03, 0.03, 0.09 - EA/ROW: 0.03, 0.03, 0.03, 0.09 - UK: 0.08, 0.08, 0.08, 0.23 - Spain: 0.06, 0.06, 0.06, 0.19 - France: 0.05, 0.05, 0.05, 0.14 - US: 0.02, 0.02, 0.02, 0.06 - Turkey: 0.17, 0.05, 0.29, 0.51 - Adverse 3 Year totals (2019, 2020, 2021, Total 3 Year): - Austria: 0.36, 0.46, 0.57, 1.38 - Czech Republic: 0.48, 0.48, 0.48, 1.44 - Slovakia: 0.37, 0.38, 0.69, 1.43 - Hungary: 0.98, 1.70, 2.42, 5.10 - Romania: 0.56, 0.89, 1.22, 2.67 - Bulgaria: 0.76, 1.16, 1.56, 3.48 - Slovenia: 0.29, 0.29, 0.29, 0.87 - Croatia: 0.95, 1.49, 2.04, 4.47 - Poland: 0.53, 0.53, 0.53, 1.58 - Russia: 0.74, 0.74, 0.74, 2.21 - Ukraine: 0.47, 0.47, 0.47, 1.40 - Germany: 0.14, 0.19, 0.24, 0.57 - EA/ROW: 0.35, 0.42, 0.50, 1.26 - UK: 0.23, 0.37, 0.50, 1.10 - Spain: 0.41, 0.54, 0.66, 1.61 - France: 0.40, 0.64, 0.88, 1.92 - US: 0.18, 0.50, 0.81, 1.49 - Turkey: 0.70, 0.70, 0.70, 2.11 ### Translation of 3-year projections into yearly projections - The structural model produces 3-year scenario-dependent loss rates. - To produce yearly projections, the model is sequentially run using the 1-year, 2-year and 3-year scenario loss rate projection. - Each run uses the 1-year loss rate projected by the previous run to infer the annual loss rate corresponding to the cumulative loss rate for the total number of years and the scenario at that point in time. - Result: 3-year loss rate projections are translated into yearly projections with cumulative impact anchored to the original 3-year loss rate projection and end-horizon scenario (presented in Table IV.1 for all material geographies). ### Bank-specific translation and starting point adjustment - Model projections for both baseline and adverse are translated into bank-specific projections using the bank mortgage exposure starting points. - Translation uses an absolute shift in the PD space to ensure the demographic component of the model is correctly captured in the baseline scenario projection. - The starting point adjustment aligns baseline scenario projections closer to idiosyncratic default rates actually observed in the market under current conditions. - Adverse scenario loss rates are projected as an additional delta impact versus the baseline. ### Overall assessment of projected loss rates - Overall the loss rates projected by the model are relatively moderate for most countries (Table IV.1). - Reasons: - Very contained population in terms of starting DSR. - Moderate house price shocks for most countries in the adverse scenario. - The overall stock of mortgage exposures is heavily concentrated in safer partitions of the DSR/LTV space (low starting current DSR and relatively low LTVs). - A more pronounced loss rate in the adverse scenario would require a combination of much higher [positive] interest shocks and negative house price developments. - Empirical illustration: countries with high 3-year cumulative loss rates in the adverse scenario are associated with a strong positive interest rate shock (Hungary, Romania, Bulgaria, and Croatia). ### Annex V — Foreign Currency (FC) and Repayment Vehicle (RV) Loans: scope and purpose - ARNIE’s FC/RV loan module quantifies loss potential arising from: - Appreciations of FC loan denomination currencies. - Underperformances of repayment vehicles attached to bullet loans. - It covers indirect credit risk as an add-on to macro-economic credit risk in the solvency stress test. - Different methodology applied to Austria and CESEE exposures due to: - Different loan characteristics (domestic FCLs mainly bullet loans; CESEE FCLs mainly installment loans). - Different data sources. - Coverage: - Austria: RV loans in EUR, CHF and JPY. - CESEE: CHF, EUR (in non-EUR countries) and USD loans. - Includes private household and corporate exposures. ### FC/RV scenarios and assumptions - Baseline and adverse FX paths from IMF macro-economic scenario are used. - In baseline, appreciation of relevant FCs is limited. - In adverse, severe appreciations in most cases: CHF appreciates versus the EUR by around 50 percent; EUR appreciates versus local CESEE currencies in the range of 10 to 30 percent. - Performance of RVs assumptions: - Baseline: 2 percent p.a. (nominal). - Adverse: 0. ### Domestic FC and RV loans methodology - All domestic RV loans (most of which are FC) considered. - Domestic loans are mostly bullet loans and hardly show default events; troubled FC loans often converted to Euro. - Method is indirect, extending former FSAP methods for FC loans to include RV loans. - Key equivalence: an RV shortfall at maturity is treated as equivalent to an FC appreciation. - Core model relationships: - ΔI = – D/I * ΔFX. (1) - D = individual debtor FC debt in EUR p.a.; I = disposable income (after debt) in EUR p.a.; ΔFX = appreciation of loan currency. - ΔLLPR = f (ΔGDP). (2) - Relationship between yearly changes of LLP ratios of loans denominated in euros and yearly changes of real GDP (proxy for ΔI). - With linear f(.) and estimated slope β = -2.5: - ΔLLPR = 2.5 * D/I * ΔFX. (3) - Application details: - Methodology applied to OeNB’s integrated reporting framework data, which contains information on remaining maturities of FC loans. - Additional impairments from (3) distributed equally over the loan’s remaining maturity; share allotted to the first 3 years represents the scenario loss. - In (3) the ratio D/I (debtor “leverage”) is set to 2.5 based on supervisory minimum standards. - ΔFX is measured with respect to the starting point of the loan, not the stress-test starting point. ### FC loans in CESEE methodology - Historical data available on LLP stocks broken down by currencies since 2007. - The excess growth of LLP ratios in FC over LC growth (the “boost-factor”) is related to historically observed FC appreciations. - Relationship assumed non-linear; exponential and quadratic functions used with different estimation criteria (quadratic, absolute, Huber-type). - This produces 15 different models; the average is taken. - Models calibrated on CHF used for EUR and USD due to better fit. - Additional credit risk costs calculated as the difference of LLP increases with and without the boost factor. ### Annex VI — LCR Scenarios: key assumptions (Table VI.1) - Scenario run-off rates and change in liquid asset weights (selected parameters): - stable retail deposits: Regulatory LCR 5%10.0%5.0%10% (table entries presented as in source) - other retail deposits: 10%20.0%10.0%20% - operational deposits: 5-25%10-35%15-35%15-35% - non-operational deposits: 20-40%20-40%40-60%40-60% - level 1 assets: nonono-5/0% - level 1 covered bonds: nonono-20/-3% - level 2A assets: nonono-15/-5% - level 2B assets: nonono-25/-5% ### Annex VII — Cash Flow-based Stress Test: haircuts and run-off rates (Table VII.1) - Scenario run-off rates, roll-off rates, haircuts, and market price shock (selected entries): - Parameter positions and values (Mild, Moderate, Severe): - stable retail deposits: 4.9% , 8.0% , 10% - other retail deposits: 9.7% , 16.1% , 20% - deposits from credit institutions: 48.7% , 80.3% , 100% - deposits from corporates: 14.5% , 27.1% , 35% - retail and corporate customers: 0.0% , 0.0% , 0% - credit intitutions: 48.7% , 80.3% , 100% - level 1 assets: 2.2-3.3% , 2.4-4.3% , 2.5-5% - level 2A assets: 8.2-16.3% , 12.4-24.7% , 15-30% - level 2B assets: 8.2-29.5% , 12.4-29.7% , 15-50% - level 1 assets (price shock): -0.2% , -0.9% , -5% - level 2A assets (price shock): -0.2/-0.6% , -0.9/-2.8% , -5/-15% - level 2B assets (price shock): -1% , -5% , -25% ### Annex VIII — The Network Model: contagion mechanics and parameterization - Balance sheet identity for bank i in a network of N banks (conceptual form presented). - Near-default of bank h causes credit losses (휆휆푥푥 ℎ푖푖) in bank i; identity adjusted accordingly. - Asset sales at discount and loss of funding incorporated: discount rate 훿훿 and fraction of funding not replaceable 휌휌. - Failure criterion: - A bank fails if its total capital adequacy ratio is below threshold 4.5 percent, triggering contagion rounds. - Parameters set in the analysis: - loss given default ratio 휆휆 = 0.5 for unsecured interbank exposure - loss given default ratio 휆휆 = 0.2 for secured interbank exposure - loss of funding ratio 휌휌 = 0.35 - discount rate 훿훿 = 0.3 *International Monetary Fund — Excerpted content as provided.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2020/english/1autea2020006.pdf_