## EXECUTIVE SUMMARY ## Source details **Canonical URL:** [EXECUTIVE SUMMARY](https://www.imf.org/-/media/files/publications/cr/2023/english/1islea2023003.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2023/english/1islea2023003.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2023/english/1islea2023003.pdf.json) --- ### INTRODUCTION - Scope and framework: - FSAP coverage situates the financial and banking system and presents selected banking indicators, bank business models, corporate sector development, risks to household debt service, and the stress testing framework. - Stress testing components: - Top-down solvency stress test of banks, macro-financial feedback loops, liquidity risk analyses (LCR, cashflow, NSFR), interconnectedness and contagion analyses, systemic liquidity analysis, pension fund stress test, non-financial corporate stress test. ### Financial sector overview - Structural transformation and size metrics: - Banking sector assets reduced from ten times of GDP to 410 percent of GDP from 2007 to 2022Q3. - Total financial sector assets reached 410 percent of GDP in September 2022. - Banking sector assets contracted from 10 times of GDP in 2007 to about 135 percent of GDP by 2022Q3. - Pension funds assets at 176 percent of GDP as of end-2022 and account for 42 percent of total financial system assets. - Macroeconomic risks: - Intensified global fragmentation, tightening financial conditions, and volatile market sentiment amplify downside risks for the Icelandic economy and financial sector. ### Banking sector soundness and vulnerabilities - Soundness indicators (as of 2022Q3): - CAR: 23 percent. - T1: 21 percent. - CET1: 20 percent. - LCR increased from 151 to 210 percent in 2022Q3. - Non-performing loans contained to below 2 percent. - Credit-to-GDP at 157 percent, corresponding to a negative 16 percent credit-to-GDP gap as of end-2022. - Profitability drivers: - High interest margin, low provisions, high fees and commissions, low cost-to-asset ratio. - Notable vulnerabilities: - FX funding from abroad ~25 percent of bank liabilities. - CRE exposure above 20 percent of total corporate loans. - Pension funds and other NBFIs hold large public debt holdings and have close ties with banks. - Continued issuance of unsecured long-term subordinated debt instruments to fulfill MREL requirements is vital. ### Bank funding and pension exposure - FX bond maturities and funding shifts: - 16 percent (or 130 billion Krona) of FX bond will mature in 2023. - 23 percent (or 185 billion Krona) will mature in 2024. - Banks increasingly shifted to covered bond issuance to meet upcoming rollover needs of uncovered FX bonds. - Pension fund funding and holdings: - Pension funds provide material funding to banks at 12 percent of total bank funding and are major shareholders of two of the three D-SIBs. - Pension funds could redirect investment offshore if foreign exposure limits permit. - Inflation indexation (as of 2022Q3): - Roughly 22 percent of total loans are indexed. - 17 percent of total liabilities (deposit and bond issued) are indexed. - Banks generally keep a positive net inflation indexed position. ### Credit risk, RWA profile, and asset composition (as of 2022Q3) - RWA and asset composition: - Credit risk RWAs account for 84 percent of total RWAs. - RWA density now just below 64 percent. - Loans: 80 percent of assets. - Central bank reserves and non-interest earning assets: 8.3 percent of assets. - Loan concentration: - Households: 51 percent of loans. - Corporates: 44 percent of loans. - Foreign loans: around 1 percent of total loans. - NPL ratios: - Corporate NPLs: 2.5 percent. - Household NPLs: 0.7 percent. - Accommodation and food services NPL ratio: 7.2 percent. - Forbearance: - 57 percent of total NPLs re-negotiated and restructured (household 38 percent; corporate 62 percent). - Marketable securities and duration: - Debt securities holdings: 384 billion Krona (8 percent of total banking assets). - Average duration: around 3.4 years. - Sovereign securities: 6.6 percent of total assets (5.4 percent domestic sovereign). ### Solvency stress test results — baseline and adverse - Sample coverage: - Three D-SIBs covering around 95 percent of banking sector assets: Arion bank 28 percent, Islandsbanki 31 percent, Landsbankinn 36 percent. - Baseline scenario: - Fully loaded CET1 ratios slightly increase from 20 to 21 percent (dividends at 50 percent). - Adverse scenario (5-year horizon and trough): - Aggregate declines by 5th year: CAR, T1 and CET1 decline respectively by about 4.2, 3.8 and 3.6 percentage points. - At the trough: CAR, T1 and CET1 decline by 5.6, 5.3 and 5.2 percentage points, respectively. - Main contributors to 5-year cumulative decline: credit risk provisioning 5.1 percentage points; RWA change 4.1 percentage points; interest rate risk 2.1 percentage points. - Market risk minimal: reduction of 60 basis points of CET1 at trough and almost 0 basis points by end of horizon. - Loss provisions: household portfolio 2.2 percent of RWA; corporate portfolio 2.5 percent of RWA over the risk horizon. - Leverage ratios: start at 14 percent, decline to 11.5 percent at the trough; minimum regulatory threshold at 3 percent. - Heterogeneity: - Aggregated results mask heterogeneity; some banks underperform due to weaker starting capital, higher default rates and constrained income generation. ### Macro-financial feedback loops and second-round effects - SVAR linkage and calibration: - Change in bank capital from solvency stress-test used to generate credit supply shock in Bayesian SVAR. - Adverse scenario aligned with 5 percent Growth-at-Risk estimate, implying a 13 percent (or 2.2 standard deviation) shock to real GDP growth relative to the baseline. - Adverse scenario implies a 9.3 percent decline relative to the starting point over a two-year horizon. - Second-round macro outcomes at trough (2024): - Real GDP: 3.8 percent lower (additional cumulative reduction). - Unemployment: 1.6 percentage points higher (additional cumulative increase). - Nominal house prices: 9 percent lower (additional cumulative fall). - Nominal exchange rate: 11.4 percent depreciation (additional cumulative). - Change in inflation rate: negligible. - Change in policy rate: small, magnitude of 60bps or less in every year. - Deleveraging at trough: 14 percent aggregate real loan growth contraction. - Quantitative capital effects: - Second-round aggregate CAR lower than first-round by 1.2 percentage points in 2024 and by 1.7 percentage points in 2027 (average for the three banks). - One bank falls slightly below the hurdle rate at the trough in the second round; aggregate capitalization remains above the hurdle rate. - CCyB calibration (reverse-stress-testing): - Additional 80bps of CCyB (relative to the 2 percent starting point) required such that all banks remain above hurdle rate; resulting CCyB: 2.8 percent. ### PD and LGD projection methodology - PD projections: - Household and corporate PDs estimated separately using Bayesian Model Averaging (BMA) with historical default rates provided by the CBI. - Conditional PD forecasts: PDs flat in baseline; sharply increase in adverse scenario; corporate impact larger than household. - LGD methodology: - Mortgage LGD structural approach (Gross et al., 2020) using LTV and house price path. - Indexed mortgages: LTV determined by house price growth and inflation (indexed principal). - Non-mortgage LGDs projected using historical coverage ratio series. ### Interest rate risk, repricing and net interest margin effects - Repricing and rates (as of 2022Q3): - Net interest margin: around 3 percent. - Mortgage lending rates: indexed mortgages 2.3; non-indexed mortgages 6.3. - Deposit rates: indexed deposits around 0.1 percent; non-indexed deposits 3.5 percent. - Household floating rate loans: about 52 percent of household loans. - Roughly 70 percent of loans repricing within less than a year. - 94 percent of deposits repricing within less than a year. - Satellite model findings: - Liability side: overnight deposit cost largely determined by short-term rate; term deposits mainly driven by long-term interest rate. - Lending rates largely explained by long-term interest rate and pass-through from funding costs. - Net interest margin impact: in adverse scenario, NIM declines by about one percentage point on average (starting NIM around 3 percent). ### Sensitivity analysis — interest-rate shock and concentration risk - 2 percentage points parallel yield curve upward shock (additional to initial adverse): - Capital ratios see an additional 250 basis points decline relative to the initial adverse by end of horizon. - Two banks face challenges meeting hurdle rates. - Drivers: higher market risk (short term), higher interest rate risk (NII effect), higher credit risk (short/medium term). - Concentration risk (default of top 5 exposures): - LGD 40 percent → CET1 depletion 2.4 percentage points; no banks breach hurdle rates. - LGD 60 percent → CET1 depletion 4.7 percentage points; no banks breach hurdle rates. - LGD 100 percent (zero recovery) → aggregate CET1 decline of 9.3 percentage points (from 19.9 to 10.6 percent); under this assumption two out of three D-SIBs may struggle to meet regulatory minimum CET1. ### Liquidity risks and stress testing (LCR, cashflow, NSFR) - Liquidity stress test design: - Three tests: LCR-based (30-day horizon), cashflow-based (12-month horizon), NSFR-based (structural; 100 percent limit). - Data sources: COREP reports as of Oct-2022. - Key liquidity metrics and positions: - Loan-to-deposit ratio: 150 percent. - 44 percent of funding placed within the overnight bucket. - Over 85 percent of cash inflows materialize beyond the first three months. - Aggregated asset encumbrance ratio: around 15 percent. - LCR scenarios and selected parameters: - Six scenarios (S1–S6) ranging from standard LCR to severe pension and foreign funding outflows; example parameters preserved (see source). - HQLA haircuts in severe scenarios include negative adjustments informed by market-value declines and ECB valuation haircuts. - LCR results: - Starting aggregate LCR: 200 percent as of 2022Q3. - Under most severe combo scenario (retail + wholesale) aggregate LCR falls to 122 percent; one bank marginally below minimum among three D-SIBs. - Adding pension and foreign funding outflows reduces aggregate LCR to around 76 percent; one more bank breaches minimum. - Currency-level: under existing regulatory minima (50 percent ISK, 80 percent EUR, no USD limit), one bank breaches threshold in USD; all D-SIBs meet EUR and ISK minima. If 100 percent applied to all significant currencies, one bank breaches EUR and one breaches USD; none of the D-SIBs would meet ISK minimum under most severe scenario. - Cashflow-based stress test findings: - System-level 12-month cash shortfall under most severe scenario: 283 billion (roughly 6 percent of total assets). - Over 30-day horizon, banks on aggregate can broadly withstand outflows supported by counterbalancing capacity; two banks experience liquidity shortfalls within 30 days under most severe scenario (bank-specific shortfalls over 12 months: 348 billion, 7.3 percent of total assets). - Currency-specific bank-level shortfalls over 30-days (12-months): - Euro: 0.03 (0.4) billion = 0.08 (1.3) percent of total banking assets. - U.S. dollar: 0.16 (0.2) billion = 0.5 (0.6) percent of total banking assets. - Icelandic Krona: 108 (234) billion = 2.3 (4.9) percent of total banking assets. - NSFR results (as of 2022Q3): - Aggregate NSFR: 123 percent. - NSFR by currency: Euro ~264 percent; Total currency 123 percent; Icelandic Krona 111 percent; US dollar ~93 percent. - NSFR stress migration results: no bank below 100 percent for total currency; one bank failed for Euro, one for US dollar, two for Icelandic Krona. - Policy implications: - Enhance monitoring of LCR by currencies and develop approaches to monitor funding risks from NBFIs (including pension funds) and foreign investors. - Consider separate LCR requirements at 100 percent for individual significant currencies; address outlier banks via Pillar 2 and supervisory actions. ### Systemic FX liquidity analysis and central bank backstop - Stress test design: - Simulates joint outflows from households, NFCs, nonresident deposits, international bond maturities, and funding shocks from pension and other NBFIs; 30 percent currency depreciation assumed in adverse scenarios; local currency conversion allowed in some scenarios. - Shock parameter ranges (preserved exactly): - Shock 1 HH FX deposit shock: [10,20]; Local Currency Conversion Y [5,10] - Shock 2 NFC FX deposit shock: [20,40]; Local Currency Conversion Y [10,20] - Shock 3 Nonresident deposit shock: [20,80]; Local Currency Conversion N/A - Shock 4 International bond outflows: [15,30]; Local Currency Conversion N/A - Shock 5 Funding shock from pension funds and other NBFIs: [20,40]; Local Currency Conversion Y [10,20] - Shock 6 FX liquid asset haircut: [2,20]; Local Currency Conversion N/A - Results and reserves: - Maximum FX liquidity gap under severe scenario: 3.4 billion US dollars (about half of gross international reserves). - Gross international reserves: around 6.5 billion US dollar (as of end-2021). - Net international reserves (netting FX borrowings): roughly 3.7 billion US dollar. - Of total FX government bonds outstanding: 500 million Euro matures in 2024; 500 million Euro in 2026; 750 million Euro in 2027. - Under severe scenarios, gross and net reserves remain positive; identified gaps can be met by swap agreements with European banks or direct CBI FX liquidity support. - Single-factor sensitivity tipping points: - International bond outflows: 20–60 percent. - Non-resident FX deposit outflows: 40–100 percent. - FX funding outflows from pension funds and NBFIs: 20–40 percent with LC conversion 10–30 percent. - Banks can withstand standalone shocks up to specified limits; LC conversion for NFC above 50 percent would create FX pressure (154 million US dollars gap); LC conversion for NBFI above 30 percent yields 320 million US dollars gap. ### Pension funds and NBFIs — scale, vulnerabilities, and stress testing - Pension fund sector size and composition (exact figures preserved): - Assets almost ISK 7 trillion at end-2022, or 176 percent of GDP. - Pillar II schemes manage ISK 5.9 trillion. - Pillar III providers manage ISK 1.0 trillion. - Pillar II offered by 21 pension funds; four largest Pillar II providers hold a cumulated market share of 61 percent. - Direct stocks: 19 percent of Pillar II and 7 percent of Pillar III assets. - Around one third of Pillar II assets invested in inflation-linked fixed-income assets; mortgages account for 10 percent of assets. - Combined exposure to domestic banks, corporates and sovereign increased from ISK 250 billion in 2017 to ISK 670 billion in 2021 (6 to 10 percent of pension fund assets); corresponds to 14 percent of banks’ financial liabilities. - As of end-2022, 37 percent of Pillar II assets denominated in foreign exchange (24 percent in Pillar III), mostly in US dollar; currency risks typically not hedged. - Pension fund stress test adverse scenario (median fund impact): - Asset values decline by 13 percent in 2023 and further 3 percent in 2024, before recovering +10 percent in 2025. - Future pension values decline by between 8 and 15 percent for a member with 10 years prior to retirement (median fund decline 13 percent). - For a member 30 years prior to retirement, median fund decline: less than 1 percent (range across funds: minus 2 percent to plus 0.5 percent). - Sensitivity analyses: - Discount rate shock: lowering discount rate from 3.5 percent to 3.0 percent increases liabilities by 15 percent and deteriorates actuarial position by almost 10 percentage points (future position -11 percentage points; accrued position -8 percentage points). - Mortality shock: decline of mortality rates by 10 percent increases liabilities by around 2 percent and reduces actuarial position by 2 percentage points. - Liquidity risks and flows: - Pillar II: accumulating funds; predictable flows; large funds held no derivatives or off-balance instruments as of end-2022. - Pillar III: cash flows more volatile; extraordinary withdrawals in crises (total withdrawals 2020: ISK 23 billion; 2021: ISK 13 billion); quarterly outflows have not exceeded ISK 100 million for funds that experienced outflows. - Pension funds hold around 22.7 percent share in mortgage lending market as of end-2022; mean LTV for newly issued loans ~50 percent; mean DSTI slightly above 20 percent. - Concentration to banks: - Exposures to Icelandic banks ~10 percent of pension fund assets. - Exposures to banks correspond to 14 percent of banks’ financial liabilities. - For each of two largest banks, exposure averages more than 3 percent of pension fund assets; exposure to whole domestic banking sector slightly more than 10 percent (45 percent covered bonds; 38 percent shares; 14 percent deposits). - Recommendations for pension funds and oversight: - Improve supervisory reporting quality and automated validation rules. - Closely monitor effects of higher inflation and interest rates on pension funds’ investment behavior and counterparty default risk. - Analyze pension funds’ mortgage lending practices (pricing, provisioning, risk management). - Focus liquidity analysis on smaller funds and volatile Pillar III cash flows. ### Non-financial corporate stress testing and firm-level vulnerabilities - Data and coverage: - ORBIS data covering ~24,000 non-financial companies (2020 vintage); NFC sample represents about 70 percent of active firms. - Key findings: - Median firm in 2020: constrained ICR, low ROA, high debt ratio despite reasonable cash coverage. - Interest expense shock: 30 percent increase in interest expenses raises share of firms at risk by 5 percent and debt at risk by 13 percent. - Scenario results: - Baseline: share of debt for firms with ICR<1 increases from 43 percent to about 70 percent within a year; probability of default increases by about one percent. - Adverse: share of debt for firms with ICR<1 rises to 96 percent in first year; probability of default peaks at 14.3 percent in first year. ### Interconnectedness, contagion and market spillovers - Domestic interbank and bank–NBFI networks: - Three commercial banks account for 95 percent of total banking assets; domestic interbank exposures small relative to capitalization. - Contagion through credit and funding channels very limited; no single domestic bank failure would trigger failure of others in tests. - Bank–NBFI network: pension funds primary creditors (80 percent of asset exposures to banks); banks primary borrowers (94 percent of liabilities in Bank–NBFI system). - Banks most contagious under credit shocks; pension and investment funds more vulnerable in some instances but overall vulnerability under 2.5 percent. - Cross-border contagion: - Icelandic banks’ net foreign asset position for banks: aggregate negative ~13 percent of GDP (about 60 percent of capital); central bank international reserves and pension funds’ foreign assets turn financial system net foreign position positive. - Cross-border exposures geographically clustered: ~two-thirds Europe; ~30 percent U.S. and Canada. - Espinosa-Vega and Sole (2010) parameters for cross-border test: lambda 70 percent, delta 50 percent, rho 35 percent. - Icelandic banks not a source of contagion to major economies but vulnerable to inward spillovers; inward spillover from U.S. most significant among sample countries. - Market-based spillovers: - Diebold and Yilmaz (2014) analysis shows Iceland is a net receiver of stock return and sovereign CDS return spillovers; primary equity return contributors: France, the Netherlands, Germany; Italy primary source of net return connectedness for sample. ### Key policy recommendations (authorities and timeline) - Systemic risk analysis and monitoring: - Develop stress testing approaches to monitor funding risks from NBFIs (including pension funds) and foreign investors. - Authorities: MoFEA/CBI; Timeline: MT. - Differentiate inflation-indexed and non-indexed lending and funding instruments in analysis of inflation impact on banks’ credit, interest rate, and market risks. - Authorities: MoFEA/CBI; Timeline: MT. - Continue liquidity stress tests with various runoff and haircut rates, enhance monitoring of LCR by currencies, and address outlier banks through Pillar 2 and supervisory actions. - Authority: CBI; Timeline: NT. - Perform data quality checks for banks and pension funds’ supervisory reporting, require corrections, and expand automated validation rules. - Authority: CBI (FSA); Timeline: NT. - Closely monitor impact of higher inflation and interest rates on banks’ solvency and pension funds’ behavior, counterparty default risk, and Pillar III cash flows for smaller funds. - Authority: CBI (FSA); Timeline: NT. - Analyze pension funds’ mortgage lending practices (pricing, provisioning, risk management). - Authority: CBI; Timeline: MT. - Timeline legend: - I = Immediate (within one year); NT = Near Term (within 1-3 years); MT = Medium Term (within 3-5 years). *Source: EXECUTIVE SUMMARY, 1islea2023003.* ### EXECUTIVE SUMMARY _________________________________________________________________________7 ### EXECUTIVE SUMMARY _________________________________________________________________________7 ### INTRODUCTION - Identifies the scope of the FSAP coverage and situates the financial and banking system (Figure 1). - Presents selected banking indicators and bank business models (Figures 2–3). - Discusses corporate sector development and risks to household debt service (Figures 4–5). - Presents the stress testing framework (Figure 6). ### TOP-DOWN SOLVENCY STRESS TEST OF BANKS - A. Banking Sector Vulnerabilities by Risk Categories (page 22) - Describes risk categories and shows Risk Weighted Assets and Density (Figure 7). - Presents bank asset composition and credit quality (Figure 8). - Breaks down debt securities and residual duration (Figure 9). - B. Solvency Stress Tests of the Banking Sector (page 26) - Details scenario design and output contractions (Figures 10–11). - Presents the structure of the solvency stress testing framework (Figure 12). - Shows PD Projections and Interest Rate Projections (Figures 13–14). - Reports results for bank solvency stress test and interest rate sensitivity analysis (Figures 15–16). - Includes corporate sector sensitivity analysis (Figure 17). - C. Recommendations (page 42) - Policy recommendations are contained in the recommendations subsection. ### MACRO-FINANCIAL FEEDBACK LOOPS - Discusses second-round scenario analysis and calibration tools (Figure 18; Box 1: Calibration of the CCyB using the Macro-Financial Linkages Model). ### BANKS LIQUIDITY RISK ANALYSIS - A. Introduction (page 48) - B. Liquid Assets and Funding Structure (page 48) - Presents bank liquidity position and time-bucket compositions (Figures 19–20). - C. LCR-Based Liquidity Stress Test (page 51) - LCR stress test parameters summarized (Table 2). - D. Cashflow-Based Liquidity Stress Test (page 54) - Cashflow stress test parameters summarized (Table 3). - E. NSFR-Based Liquidity Stress Test (page 58) - NSFR development and methodology (Figures 23–24). - F. Recommendations (page 60) ### BANK INTERCONNECTEDNESS ANALYSIS - A. Domestic Cross-Sectoral Interconnectedness (page 61) - Balance sheet exposures and methodology (Figure 25). - B. Domestic Interbank and Bank-NBFI Network (page 63) - Domestic systemic interconnectedness test results (Figure 26). - C. Cross-Border Interbank Contagion (page 66) - Cross-border interconnectedness test results and sectoral breakdowns (Figures 27–28). - D. Market-Based Interconnectedness Analysis (page 69) - Gross and net contributions to systemic risk (Figure 29). ### SYSTEMIC LIQUIDITY ANALYSIS - A. Introduction (page 71) - B. Overview of Sectoral Balance Sheet and Systemic Liquidity Linkages (page 72) - Stylized balance sheet by sector (Figure 30). - C. Methodology (page 74) - Transmission channels and shock parameters (Figure 31). - D. Results (page 76) - Results for systemic liquidity stress test (Figure 32). - E. Sensitivity Analysis (page 78) - Single factor sensitivity analysis (Figure 33). - F. Recommendations (page 80) ### PENSION FUND STRESS TEST - A. Scope and Sample of the Analysis (page 80) - Pension fund market and asset allocation (Figure 34). - Pension fund sample details (Table 4). - B. Scenario and Methodology (page 83) - Adverse scenario for pension funds (Table 5). - C. Impact on Future Pension Values (page 85) - Impact on pension fund assets and future pension values (Figures 35–36). - D. Sensitivity Analyses (page 86) - Pension fund sensitivity analyses (Figure 37). - E. Liquidity Risks (page 87) - Pension fund liquidity risks and mortgage lending exposures (Figures 38–39). - F. Further Vulnerabilities (page 89) - Pension fund exposures to domestic banks and annual investment flows (Figures 40–41). - G. Recommendations (page 92) ### NON-FINANCIAL CORPORATE STRESS TEST - Presents corporate sector vulnerabilities and scenario analyses (Figures 42–43 and Figures 42–43 for sensitivity and scenario analysis). ### FIGURES AND TABLES (selected) - Figures 1–43 present the analytical evidence, including stress test frameworks, results, and sensitivity analyses. - Tables include: Table 1 (2023 FSAP: Key Recommendations), Table 2 (LCR Stress Test Parameters), Table 3 (Cashflow Stress Test Parameters), Table 4 (Pension Fund Sample), Table 5 (Adverse Scenario for Pension Funds). ### APPENDICES (selected) - I. Capital Flow Management Measures in the Icelandic Toolkit (page 96) - II. Risk Assessment Matrix (page 97) - III. Growth at Risk (page 99) - IV. Macro-Financial Linkages (page 103) - V. The Estimation of the PD Satellite Models (page 106) - VI. The Estimation of the Interest Rate Satellite Models (page 107) - VII. Financial Soundness Indicators—A Regional Comparison (page 109) - VIII–XI. Stress Testing Matrices for Banking, Interconnectedness, Pension Fund, and Non-Financial Corporate sectors (pages 110–118) ### GLOSSARY (selected acronyms and terms) - AC Amortized Cost - AE Asset Encumbrance - ASF Available Stable Funding - BIS Bank for International Settlements - BPS Basis Points - BSCR Basic Solvency Capital Requirement - CAR Capital Adequacy Ratio - CB Central Bank - CBI Central Bank of Iceland - CCyB Countercyclical Buffer - CET1 Core Equity Tier 1 - CFLST Cash Flow-based Liquidity Stress Test - FSAP Financial Sector Assessment Program - FX Foreign Exchange - GDP Gross Domestic Product - HQLA High-Quality Liquid Assets - ISK Icelandic Krona - LCR Liquidity Coverage Ratio - LGD Loss Given Default - LTV Loan-To-Value - NBFI Non-Bank Financial Institution - NFC Non-Financial Corporate - NPL Non-Performing Loan - NSFR Net-Stable Funding Ratio - PD Probability of Default - RWA Risk-Weighted Assets - RWD Risk Weight Density - SME Small- and Medium-Sized Enterprises - ST Stress Test - TD Top-Down (stress test) *Source: EXECUTIVE SUMMARY, 1islea2023003* ### EXECUTIVE SUMMARY ### EXECUTIVE SUMMARY ### Financial sector overview - The Icelandic financial landscape has undergone significant structural transformation since the global financial crisis (GFC), with a contracted banking sector and growing role of pension funds. - Asset and size metrics: - Banking sector assets reduced from ten times of GDP to 410 percent of GDP from 2007 to 2022Q3. - Total financial sector assets reached 410 percent of GDP in September 2022. - Banking sector assets contracted from 10 times of GDP in 2007 to about 135 percent of GDP by 2022Q3. - Pension funds have assets at 176 percent of GDP as of end-2022 and account for 42 percent of total financial system assets. - The financial system weathered the global pandemic on the back of strong fundamentals, but intensified global fragmentation, tightening financial conditions, and volatile market sentiment amplify downside risks for the Icelandic economy and financial sector. ### Banking sector soundness and vulnerabilities - Soundness indicators: - CAR, T1 and CET1 ratio stand at 23, 21 and 20 percent as of 2022Q3. - LCR increased from 151 to 210 percent in 2022Q3. - Non-performing loans contained to below 2 percent. - Credit-to-GDP at 157 percent, corresponding to a negative 16 percent credit-to-GDP gap as of end-2022. - Profitability drivers: high interest margin, low provisions, high fees and commissions, and low cost-to-asset ratio. - Notable vulnerabilities: - FX funding from abroad still amounts to about 25 percent of bank liabilities. - High exposure to commercial real estate sector at above 20 percent of total corporate loans. - Banks increasingly shifting to covered bonds issuance amid rising FX funding spreads; continuing issuance of unsecured long-term subordinated debt instruments to fulfill MREL requirements is vital. - Pension funds and other nonbank financial institutions (NBFIs) hold large holdings of public debt and have close ties to the banking system. ### Stress test results — solvency - Baseline scenario: - Banks’ fully loaded CET1 ratios sustain at a high level, slightly increasing from 20 to 21 percent, driven by strong profitability partially offset by dividend distribution policy at 50 percent. - Adverse scenario: - Aggregate declines in capital ratios by the 5th year: CAR, T1 and CET1 decline respectively by about 4.2, 3.8 and 3.6 percentage points. - At the trough: CAR, T1 and CET1 decline by 5.6, 5.3 and 5.2 percentage points, respectively. - Main contributors to declines (cumulative over 5 years): credit risk provisioning 5.1 percentage points, risk weighted assets (RWA) 4.1 percentage points, interest rate risk 2.1 percentage points. - Market risk contribution minimal: reduction of 60 basis points of CET1 at trough and almost 0 basis points by end of horizon. - Non-financial corporates incur higher credit losses than households. - Aggregated results mask heterogeneity: certain banks underperform due to weaker initial capital and higher default rates and constrained income capacity. - Macro-financial feedback loops: - A contraction in bank credit leading to additional output losses further deteriorates macro conditions, translating into an additional 1.7 percentage point CET1 ratio decline on aggregate over a 5-year horizon. - In this severely adverse scenario with feedback loops, aggregate capitalization remains above the hurdle rate, but one bank falls below the hurdle rate with a capital shortfall in CET1 amounting to 0.3 percent of GDP. - Hurdle rates context: - Under the adverse scenario, hurdle rates for CET1, Tier 1 and total capital ratio are set at minimum ratios (4.5, 6 and 8 percent, respectively) plus SRB, O-SII and Pillar II buffer. Banks are allowed to deplete CCyB and CCoB under the adverse scenario. Under the baseline scenario, hurdle rates include CCyB and CCoB. ### Liquidity risks and stress testing (LCR and cashflow) - Aggregate LCR-based stress test: - Starting point of 200 percent as of 2022Q3 declines to 122 percent in the most severe scenario combining retail and wholesale shocks; one bank saw its LCR marginally below the minimum threshold among the three D-SIBs. - When assuming additional liquidity outflow from pension and foreign funding, one more bank breached the minimum threshold and aggregated LCR fell to around 76 percent. - Cashflow-based liquidity stress test: - Banks can withstand short-term outflows using counterbalancing capacities in the short-term, but liquidity positions become much weaker beyond 30 days due to maturity mismatch (backloaded inflows and frontloaded outflows). - Most unsecured and covered bonds become due beyond 3 months. - Two banks appear to experience liquidity shortfalls within 30 days under the most severe scenario. - Currency-specific vulnerabilities: - LCR and cashflow exercises applied to significant currencies show vulnerabilities particularly in U.S. dollars and Icelandic Krona. - Under existing regulatory minimums (50 percent for Icelandic Krona, 80 percent for Euro and assuming 100 for US dollars) one bank breaches the threshold in U.S. dollars. All D-SIBs meet minimum threshold in Euro and Icelandic Krona under those assumptions. - If a 100 percent requirement were assumed for all three significant currencies, one bank breaches the minimum LCR threshold in Euro and one in U.S. dollars. - None of the D-SIBs would be able to meet minimum thresholds in Icelandic Krona under the most severe scenario. - Cashflow-based stress test: one bank would experience liquidity shortfall in Euros while two banks experience shortfalls in US dollars and Icelandic Krona separately under the most severe scenario, over both short (less than 3 months) and longer term (beyond 3 months). ### Systemic FX liquidity and central bank backstop - Systemic liquidity stress test: - Combination of FX liquidity shocks from households, NFCs, nonresident funding, pension and investment fund and international bond funding outflows identifies meaningful FX liquidity shortfalls of the banking sector, indicating the need to further build up FX liquidity buffers at banks. - If pension funds or other NBFIs move more domestic assets offshore, banks may face additional FX outflows enlarging the initial shortfall. - Identified gaps can be fully met either by swap agreements with other European banks or direct central bank FX liquidity support, given the current high level of international reserves at the CBI. ### Pension funds and NBFIs - Size and role: - Pension funds’ assets at 176 percent of GDP as of end-2022; large holdings of public debt and close ties with banking system. - Pension funds account for 42 percent of total financial system assets. - Scenario impacts: - For the median pension fund in the adverse scenario: asset values decline by 13 percent in 2023 and another 3 percent in 2024, before recovering in 2025 (+10 percent). - Future pension values decline by between 8 and 15 percent for a member with 10 years prior to retirement. - Most valuation impact stems from lower stock prices; in the first year Krona depreciation counterbalances the decline through increased value of FX-denominated investments. - Liquidity and withdrawals: - Mandatory Pillar II does not allow withdrawals; Pillar II sector overall is accumulating funds with contributions exceeding pension payments. - Pillar III allows transfers between funds, mortgage loan repayments deducted from contributions, and has experienced extraordinary withdrawals in past crises; net cashflows have remained positive in all quarters for almost every pension fund in the sample. - Asset-side vulnerabilities: - Pension funds hold around 22.7 percent share in the mortgage lending market as of end-2022. - Mean LTV ratios for newly issued loans fluctuate around 50 percent; mean DSTI ratios increased to slightly above 20 percent. - All large pension funds have concentrated exposures towards the three large domestic banks; exposure to whole domestic banking sector amounts to slightly more than 10 percent, mostly through (covered) bonds, shares and deposits. - Single-name concentrations remain below regulatory limits but sectoral concentration warrants close monitoring. - Sensitivities: - Pension funds are sensitive to changes in the valuation regime for liabilities, including discount rate and mortality assumptions. ### Corporate sector stress - Firm-level corporate stress test: - Identified notable increases in debt-at-risk amid further rise in interest rates. - Significant pickup in probability of defaults under a stress scenario aligned with the banking stress test. ### Key recommendations (Authorities; Timeline) - Systemic Risk Analysis - Develop stress testing approaches to monitor funding risks from nonbank financial institutions (including pension funds) and foreign investors. - Authorities: MoFEA/CBI - Timeline: MT - Differentiate inflation indexed and non-indexed lending and funding instruments in the analysis of inflation impact on banks’ credit, interest rate, and market risks. - Authorities: MoFEA/CBI - Timeline: MT - Continue conducting liquidity stress tests with various runoff and haircut rates, enhance monitoring of LCR by currencies, and address outlier banks through Pillar 2 and supervisory actions. - Authority: CBI - Timeline: NT - Perform data quality checks for banks and pension funds’ supervisory reporting data, require banks and pension funds to submit corrections, if necessary, and expand automated validation rules. - Authority: CBI (FSA) - Timeline: NT - Closely monitor the impact of higher inflation and interest rates on banks’ solvency condition and pension funds’ investment behavior, counterparty default risk, and (particularly for smaller pension funds) Pillar III cash flows. - Authority: CBI (FSA) - Timeline: NT - Analyze pension funds’ mortgage lending practices, in particular pricing, loss provisioning and risk management. - Authority: CBI - Timeline: MT - Timeline legend: - I = Immediate (within one year); NT = Near Term (within 1-3 years); MT = Medium Term (within 3-5 years). *Source: EXECUTIVE SUMMARY (1islea2023003).* ### 4. Nonetheless, there are some intrinsic risks associated with bank business models ### 4. Nonetheless, there are some intrinsic risks associated with bank business models ### Bank funding and pension exposure - Foreign funding (mainly unsecured debt securities and some nonresident deposits) accounts for about 25 percent of total funding. - It is expected that 16 percent (or 130 billion Krona) of FX bond will mature in 2023, and 23 percent (or 185 billion Krona) will mature in 2024. - Banks have increasingly shifted to covered bond issuance to meet upcoming rollover needs of uncovered FX bonds. - Pension funds provide material funding to banks at 12 percent of total bank funding and are major shareholders of two of the three D-SIBs. - Pension funds could re-direct investment from the domestic market to foreign markets if foreign currency exposure limits permit, creating funding pressure for banks. - Conversely, large foreign assets of pension funds could help stabilize the banking system if pension funds repatriate funds during stress. - Banks could be exposed to volatile non-resident deposits, evidenced by historical stress episodes. ### Inflation indexation of assets and liabilities - As of 2022Q3: - Roughly 22 percent of total loans are indexed. - 17 percent of total liabilities (deposit and bond issued) are indexed. - Banks generally keep a positive net inflation indexed position, which can lead to value gains during high inflation episodes as assets expand more than liabilities. - Credit-risk channels: - Non-indexed loans: interest rates can spike during periods of high inflation. - Indexed loans: charges real interest rates but add inflationary effect onto principal, which can erode debtors’ equity and lead to negative amortization during high inflation. - Borrower behavior: - Since the start of the pandemic, many borrowers chose non-indexed mortgages to benefit from ultra-low interest rates. - Recent evidence suggests migration from non-indexed to indexed mortgages due to persistently high inflation and tighter financial conditions, increasing debt burden on non-indexed mortgages. ### Exposure to tourism and commercial real estate (CRE) - Sectoral underperformance since the pandemic: - Tourism NPL ratio peaked at 17 percent. - CRE NPL ratio peaked at 8 percent. - Peak total corporate NPL ratio was 4.6 percent. - CRE accounts for over 20 percent of total corporate loans, indicating concentrated exposure. - Tourism rebound has begun, but full recovery may take longer and could be impeded by global fragmentation and regional conflicts. ### Collateral quality and lending standards - Coverage ratio of the banks is around 50 percent as of 2022Q3; this has decreased since 2018 due to asset price increases and tightened lending standards. - Household loan recovery rates are considered high and likely increased amid rising housing prices. - Loan-to-value (LTV) cap tightened: - From 85 percent (all borrowers) and 90 percent (first-time buyers) in 2017 - To 80 percent (all borrowers) and 85 percent (first-time buyers) in June 2022. - Collateral quality for corporate loans is considered riskier than household segment. - Potential over-valuation of real estate could reverse asset prices, lead to collateral re-valuation, and require increased loan loss provisions. ### Direct and indirect external exposures - Banking sector’s direct and indirect exposures to Russia and Ukraine are limited: - Imports of goods from Russia amounted to 0.6 percent of total imports from 2018 to 2020. - Exports of goods to Russia amounted to 1.1 percent over the same period. - Iceland’s energy is almost 100 percent renewable (75 percent hydropower, 24 percent geothermal). - Direct exposure to Credit Suisse and UBS: - Amounts to about 0.5 billion Krona, equivalent to 0.01 percent of total assets, as of Feb 2023 (includes on- and off-balance sheet exposure). - Securities classification and duration: - Icelandic banks do not hold debt securities classified under amortized cost. - Marked-to-market securities are limited at 8 percent of total assets, with short durations at around 3.4 years. - Cross-border exposures: - Counterparties with largest cross-border exposures include U.S., Belgium, and Canada. - Around two-thirds of exposures come from European countries such as Belgium, Norway, and Denmark. - U.S. and Canada account for 30 percent of total exposures. ### Pension funds: scale and links to banks - Total assets of the pension fund sector amount to 176 percent of GDP at end-2022. - Exposures and holdings: - Exposures to Icelandic banks account for 10 percent of total pension fund assets. - Exposures to banks account for 14 percent of banks’ financial liabilities. - Holdings of sovereign bonds account for 21 percent of pension fund assets (includes municipal bonds). - Pension funds’ activity in mortgage market: - Outstanding mortgage volume equals 23 percent of the outstanding mortgage volume. - Share of foreign-denominated assets in pension funds reached 35 percent of assets as of end-2022. - Note: As of March 2023, the limit on foreign investment ratio of the pension funds was lifted from 50 percent to 65 percent, with incremental implementation till 2036 (implication for future portfolio shifts). ### Corporate sector debt and pandemic effects - Non-financial corporate sector debt around 95 percent of GDP by end-2022. - Share of external debt in total corporate debt declined from about 50 percent in 2008 to 17 percent by end-2022. - NFCs remain highly dependent on loan financing; large share of non-indexed corporate loans could suffer under a prolonged high interest rate environment, pressuring debt servicing capacity and increasing bank credit risks. - Pandemic impacts: - Profitability declined for both publicly listed and privately held firms; leverage remained contained. - Firm-at-risk and debt-at-risk (ICR lower than 1.5) increased by 2.4 and 3.1 percentage points, respectively, in 2020 compared to prior year. - Aggregate non-performing loans on D-SIBs lending to NFCs increased marginally during the pandemic. ### Household debt and debt service capacity - Household debt increased marginally during the pandemic but has been on a downward trajectory since the GFC. - CBI simulations on DSTI: - Comparing DSTI at origination versus in January 2023, the share of borrowers with DSTI above 35 percent increases from about 7 to 15 percent. - Based on updated income as of January 2023, the share increases from about 7 to 9 percent only. - Real wage increases are about 7 percent higher than in 2019, which have boosted household debt service capacity. ### Stress testing and risk analysis framework - FSAP conducted comprehensive stress tests and risk analyses covering banking, corporate, pension fund and other NBFI sectors. - Banking analysis: - Scenario-based stress test of three D-SIBs covering 95 percent of banking sector assets on highest consolidation level. - Complemented by macro-financial feedback, sensitivity tests (corporate concentration, sector-specific shocks), liquidity stress tests, and contagion analysis (domestic and cross-border interbank exposures). - Systemic liquidity analysis assessed adequacy and FX liquidity buffers of the entire financial sector under simultaneous liquidity shocks. - Corporate sector stress test used granular firm-level balance sheet data to simulate corporate PDs under stress. - Pension funds analysis: - Top-down and bottom-up risk analysis. - Given a mostly defined-ambition regime in Pillar II (members bear investment risk), impacts of adverse scenarios were calculated on future pension values. - Additional sensitivity tests, liquidity risk assessment, and analysis of pension funds’ investment behavior and mortgage lending were performed. *Source: IMF staff and Central Bank of Iceland as presented in the chapter.* ### 13. Credit risk constitutes the largest risk factor for the banking system (Figure 7). As of ### 13. Credit risk constitutes the largest risk factor for the banking system (Figure 7). As of ### Credit risk and RWA profile - Credit risk RWAs account for 84 percent of total RWAs in the sample banks as of 2022Q3. - RWA density has been steadily decreasing and is now just below 64 percent. - Banks use the standardized approach, limiting room to lower RWA density further under current business models. ### Bank asset composition and credit quality (as of 2022Q3) - Asset composition by type: - Loans: 80 percent of assets. - Central bank reserves and non-interest earning assets: 8.3 percent of assets. - Loan concentration by sector: - Households: 51 percent of loans. - Corporates: 44 percent of loans. - Nonbank financial institutions, banks, central banks, and governments: each small in size. - Foreign loans: around 1 percent of total loans (considered immaterial). - Non-performing loan (NPL) ratios: - Corporate NPLs: 2.5 percent. - Household NPLs: 0.7 percent. - Within corporates, accommodation and food services NPL ratio: 7.2 percent (represents tourism sector exposure). ### Forbearance and NPL treatment - As of 2022Q3, 57 percent of total NPLs have been re-negotiated and restructured: - Household NPLs restructured: 38 percent. - Corporate NPLs restructured: 62 percent. - Forbearance can provide buffers for viable borrowers and reduce banks’ probability of default and provisioning needs, but banks should distinguish viable from non-viable borrowers. ### Interest rate risk and repricing - Net interest margin: around 3 percent as of 2022Q3 (higher than Nordic average). - Mortgage lending rates as of 2022Q3: - Indexed mortgages: 2.3. - Non-indexed mortgages: 6.3. - Deposit rates as of 2022Q3: - Indexed deposits: around 0.1 percent. - Non-indexed deposits: 3.5 percent. - Risks from indexed instruments: - Indexed instruments charge real interest rates; inflationary adjustments add to principal monthly and can cause negative amortization. - Higher share of indexed assets than indexed liabilities, competition from pension funds, and rising funding costs could lower net interest income under an interest rate shock. - Repricing structure: - Household floating rate loans: about 52 percent of household loans. - Most corporate loans are floating rate. - Roughly 70 percent of loans are subject to repricing within less than a year. - 94 percent of deposits are subject to repricing within less than a year. - Debt securities issued by banks generally have longer repricing profiles. - The repricing structure limits pressure on interest rate margins but may amplify credit risks. ### Market risk - Holdings of debt securities: - Amount: 384 billion Krona. - Share: 8 percent of total banking assets. - Average duration: around 3.4 years. - Sovereign securities: 6.6 percent of total assets, mostly domestic sovereign securities at 5.4 percent of total assets. - FX and equity risks: limited due to almost closed FX net open position and immaterial equity exposure. - Inflation indexation exposures: - Banks generally keep a positive inflation indexation balance, which can lead to profit during inflation as the inflation expansion of assets exceeds that of liabilities. - Valuation changes from CPI-linked products in a rising inflation context dominate other market risks. ### Solvency stress tests: coverage and data - Stress test covered 3 banking institutions constituting around 95 percent of total banking system assets. - The three D-SIBs and their shares of total banking sector assets: - Arion bank: 28 percent. - Islandsbanki: 31 percent. - Landsbankinn: 36 percent. - Supervisory data as of 2022Q3 used at highest consolidation level within Iceland; included COREP, FINREP, system and bank specific PDs and transition matrices provided by the CBI on a confidential basis. - Results assume no supportive policy actions taken to counteract negative impact on the banks. ### Adverse scenario design and Growth-at-Risk linkage - Adverse scenario features: - Intensification of regional conflict(s) disrupting trade and pushing up inflation. - Abrupt global slowdown or recession with synchronized sharp growth slowdown. - Monetary policy miscalibration de-anchoring inflation expectations, increasing risk premia, long-term bond yields, and corporate spreads. - Sudden correction in the domestic real estate market. - Systemic financial instability prompting insolvencies and market dislocation. - The multivariate adverse scenario is derived from a VARX framework using adverse paths of US GDP and oil price from the GFM model, and a domestic shock layer reflected in the unemployment rate. - Severity: - Aligned with 5 percent Growth-at-Risk estimate, implying a 13 percent (or 2.2 standard deviation) shock to real GDP growth relative to the baseline. - Implies a 9.3 percent decline relative to the starting point over a two-year horizon. - About 65 percent of total output loss is accounted for by external drivers. - Comparison of severity: - Cumulative decline relative to Y0 under adverse scenario: -9.3. - GFC cumulative decline relative to Y0: -10.3. - COVID cumulative decline relative to Y0: -2.7. ### Solvency stress test methodology and assumptions - Accounting and risk framework: - Balance sheet approach based on IFRS9 expected loss framework with asset stage classifications, transition matrices, and life-time expected loss calculations. - Capital metrics assessed: - Total capital adequacy ratio (CAR), Tier 1 capital (T1), Common Equity Tier 1 (CET1), and leverage ratios. - Capital requirements include minimum capital ratios plus bank specific Systemic Risk Buffer (SRB), O-SII buffer, and Pillar-II requirements where applicable. - Banks allowed to deplete capital conservation buffers (CCBs) and countercyclical capital buffer under the adverse scenario. - As of 2022Q3, the average SRB, O-SII buffer and Pillar II requirements across three sample banks are 2.8, 2 and 3.2 percent, respectively. - In March 2023, CBI announced an increase of the CCyB from 2 percent to 2.5 percent. - Risks covered: - Credit risk for all exposures, market risks, sovereign risk, and interest rate risk in the banking book. - Derivatives book not considered due to lack of granular information. - Translation of macro scenarios: - Macro scenarios translated into PDs, LGDs and interest rates via satellite models; affected growth of balance sheet items, pre-provision net income, and other components. - Shocked risk parameters drove RWAs and provisions (via IFRS9 transition matrices), asset repricing and market valuation losses. - Balance sheet growth assumption: - Quasi-static approach: balance sheet growth assumed to follow Iceland’s nominal GDP growth; floor on rate of change set at zero percent (binding in adverse scenario). - Balance sheet growth driven by inflation and FX movements according to bank exposures to inflation-indexed and FX-denominated assets and liabilities, and conversion of some off-balance sheet items to on-balance sheet exposures. - Inflation-indexed exposure assumed to grow with inflation dynamics at a slower pace to account for amortization. - Default evolution modelling: - Household and corporate PDs estimated separately using Bayesian Model Averaging (BMA) methodology with historical default rates provided by the CBI. - Risk-neutral PDs for sovereign, banks, NBFIs generated using Merton approach, translating credit spreads under stress to PDs with assumed LGD of 45 percent. - Resulting PDs used to shift transition matrices of each loan portfolio held by banks. - Logit transformation applied before BMA/OLS to address truncated default rate distribution and contain projections within 0-1 bounds. *Source: CBI and IMF staff.* ### 29. Conditional PD forecasts, which indicate larger impact on corporate than household ### 29. Conditional PD forecasts, which indicate larger impact on corporate than household ### PD projections and segmental impact - Conditional PD forecasts were generated based on the estimated model parameters (Figure 13). - Under a stable macroeconomic outlook in the baseline, the PDs in both segments are projected to remain flat in the baseline scenario and to sharply increase in the adverse scenario. - The adverse scenario impact is idiosyncratic across segments: the impact on corporate is more sizable than on household mortgages. - The magnitude of the projected PD shock under the adverse scenario is in general milder than historical stress episodes explained by structural changes in the economy since the GFC. ### Key drivers of credit risk - Unemployment rate, interest rates as well as asset prices proved to be relevant for the buildup of credit risk. - Significance of explanatory variables by segment: - Household PDs: unemployment rate and interest rate on indexed mortgages appear to be significant drivers. - Corporate PDs: unemployment rate, term premium and stock market prices are significant and jointly explain the evolution of corporate PDs. - The type and number of significant variables varies distinctly across segments, reflecting individual characteristics of their historical PDs. ### LGD projection and provisioning methodology - Point-in-time (PiT) LGD projections rely on both econometric approaches and structural models. - Mortgage segment: structural approach originated from Gross et al. (2020) designed to model LGDs for real estate-collateralized portfolios using information on loan to value ratio (LTV) and LGDs at the starting point as well as scenario path for house price growth. - Distinction between indexed and non-indexed mortgage portfolios: for indexed mortgages LTV is assumed determined by both house price growth (denominator) and inflation (numerator), since high inflation under stress could expand indexed-loan principal and push up the LTV. - Non-mortgage portfolios: LGDs projected using historical time series on coverage ratio for the total loan portfolio provided by the CBI. - Forward PiT LGD paths under baseline and adverse scenarios were attached to bank starting points to derive bank-specific PiT LGDs over the 5-year horizon. - Calculation of loan loss provisions is consistent with the accounting standard (IFRS9) and depends on the evolution of loan exposures, stage transition matrices (guided by stressed PiT PDs) and the PiT LGDs under stress. - Note on sample limitations: certain models could be subject to sample bias due to data limitation (e.g., coverage ratio with low R-squared). ### Interest rate risk assessment and projections - Interest rate risk assessment decomposed into: - Base effect: changes in interest income or expense due to changes in outstanding amounts absent interest rate shocks. - Gains or losses under stress: add-on component using gap-analysis to assess cash-flow effects from a general increase in interest rates across repricing buckets (less-than-1-year to 5-year). - Interest payments assumed to accrue only on performing exposures under both baseline and adverse scenarios; interest revenue on performing exposures calculated on gross carrying amount. - Projection inputs and mapping: - Historical aggregated interest rates and interest rate sensitive asset and liability series from the CBI. - Aggregate bank rates for new business mapped into three asset categories (household, corporate and consumer loans) and three liability categories (overnight deposits, term deposits, and debt securities). - Distinction between indexed and non-indexed portfolios on both asset and liability sides. - Lending rates projected allowing inclusion of funding rates (indexed and non-indexed term deposits) as explanatory variables to allow partial pass-through of rising funding cost onto lending rates. - Interest rate on indexed and non-indexed debt securities follows dynamics of term deposits plus spread benchmarked to historical average. - Satellite models show: - Liability side: cost of overnight deposits largely determined by short-term rate; term deposits driven mainly by long-term interest rate. - Large and significant pass-through from short-term rate and long-term sovereign yield to overnight and term deposit rates, particularly for indexed and non-indexed term deposits. - Asset side: lending rates largely explained by long-term interest rate and pass-through from funding costs; non-indexed mortgage, corporate and consumer loan rates also positively correlated with inflation. - Projections reflect portfolio characteristics: - More severe impact on long-term funding relative to short-term funding. - Shock to inflation-indexed funding is lower than non-indexed funding. - Lending rate increases more moderate for mortgages than for corporate and consumer loans. - Increase in indexed mortgage rates smaller than non-indexed mortgages (indexed mortgages do not consider inflation premium; competition from pension funds that grant almost exclusively indexed mortgages constrains banks). - Net interest margin (NIM) impact: - Relative to the baseline, net interest margin declines in the adverse scenario by about one percentage point on average for the sample banks. - Starting net interest margin of the total banking sector at the starting point at around 3 percent. ### Market risk valuation and other market exposures - Modified duration approach used to measure gains or losses in value of fixed income securities due to changes in bond yield; securities holdings by type and durations provided by the CBI. - Analysis covers FVTPL and FVOCI holdings, separated into sovereign and central bank, banks, NBFIs and corporate bonds. Rebalancing of the portfolio was not allowed throughout the horizon. - Valuation changes for FX, inflation and equity exposures assessed using net positions from CBI supervisory reports and projected market prices under baseline and adverse scenarios. - Banks are expected on average to benefit from an inflationary scenario due to positive net inflation positions leading to valuation gains. ### Income projection, taxes and distribution policy - Net income projected using all risk factors in the stress test; net profits mainly driven by gains and losses from credit risks, market risks and interest rate risks. - Other income items: - Net fee and commission income and other non-interest income projected to grow in line with real GDP growth under both scenarios. - Operational and administrative expenses assumed constant over the risk horizon. - Extraordinary income and loss assumed not to recur during the projection period. - Corporate income tax set at banks’ effective tax rate at 20 percent. - Dividend policy assumed: dividends paid at a rate of 50 percent of current period profit after taxes by banks that are making profits (only if profits are positive) and in compliance with supervisory capital requirements. - Banks are not allowed to issue new shares or make repurchases during the stress test horizon. ### Scenario-based stress test results - Baseline scenario: - Confirms banks’ strong capital positions, with mild capital accumulation. - Both retail and large international banks see fully loaded CET1 ratios remaining at a high level while slightly increasing from 20 to 21 percent. - Marginal capital accumulation driven by strong income generation offset by dividend distribution at 50 percent. - Adverse scenario: - Confirms resilience to severe yet plausible adverse shocks: no bank saw capital ratios fall below hurdle rates. - On aggregate, fully loaded CAR, T1 and CET1 ratios decline respectively by about 4.2, 3.8 and 3.6 percentage points by the 5th year and 5.6, 5.3 and 5.2 percentage points at the trough. - Among risk factors, cumulative contributions over 5 years: - Loss provisions (credit risk): 5.1 percentage points (largest contributor). - Change in risk weighted assets (RWA): 4.1 percentage points. - Interest rate risk: 2.1 percentage points. - Loss provision for household portfolio lower than corporate portfolio: 2.2 relative to 2.5 percent of RWA over the risk horizon. - Market risk contribution minimal: 60 basis points at trough and almost 0 by end of horizon. - Aggregated results mask bank heterogeneity: some banks underperform due to weaker starting capital, higher default rates, and constrained income generation. - Leverage ratios: start at 14 percent, decline to 11.5 percent under the adverse scenario at the trough, well above the minimum threshold at 3 percent. ### Sensitivity analysis - Further interest rate hikes affect solvency mainly through market risk, interest rate risk and credit risk. - Immediate impact on trading portfolios (marketable debt securities) via realized or unrealized valuation losses in FVTPL and FVOCI categories. - Securities held at amortized cost could incur market losses if liquidated at market value to meet liquidity outflows. - Market risks could be hedged using interest rate derivatives. - Over time, higher rates may increase or decrease net interest income depending on repricing timing and may increase loan defaults and loan-loss provisioning (LLP) costs. *Source: IMF staff.* ### 44. The sensitivity analysis simulating further tightening of financial conditions reveals ### 1islea2023003 - 44. The sensitivity analysis simulating further tightening of financial conditions reveals ### Interest-rate sensitivity of banks - Scenario: a 2 percentage points parallel increase along the yield curve (parallel upward shift in both short term and long-term bond rate in addition to the initial adverse scenario). - Key impacts: - Capital ratios saw an additional 250 basis points decline relative to the initial adverse by the end of the horizon. - Two banks faced challenges in meeting the hurdle rates as a result of the additional shock. - Drivers of capital depletion: - Market risk: higher contribution in the short term. - Interest rate risk (NII effect): higher contribution over the short and medium term. - Credit risk: higher contribution over the short and medium term. - Mechanisms behind the higher NII effect: - Positive net indexed position of the banks (banks hold higher share of inflation indexed assets relative to inflation indexed liabilities so the interest increase on the asset side overall is smaller relative to liabilities). - Partial passthrough of funding cost to lending rates, assuming increased lending competition faced by the banks. - Credit risk sensitivity: - Further rise in interest rates increases credit risks due to higher inflation, already quite high lending rates, and tightened financial condition (especially on non-indexed loans), typically with a lag as defaults and provisions materialize. ### Corporate sector sensitivity analyses - Two separate sensitivity exercises were conducted: 1. Concentration risk: simultaneous default of the top 5 exposures for each bank, with LGDs assumed between 40 percent and 100 percent (zero recovery). 2. Sectoral approach: assumed homogeneous PDs ranging from 10 percent to 30 percent for each sector, combined with sectoral and bank specific downturn LGD since 2014; provisions computed as PD × LGD and deducted from CET1 ratio. An alternative experiment used homogeneous LGDs of 40 percent across industries. - Results (concentration risk, default of top 5 exposures): - LGD 40 percent → CET1 ratio depletion of 2.4 percentage points; no banks breaching hurdle rates. - LGD 60 percent → CET1 ratio depletion of 4.7 percentage points; no banks breaching hurdle rates. - LGD 100 percent (zero-recovery) → aggregate CET1 ratio decline of 9.3 percentage points, from 19.9 percent to 10.6 percent; under this assumption two out of three D-SIBs may have trouble meeting the regulatory minimum CET1 capital. - Results (sectoral PD shocks): - Real estate activities produce the highest capital impact given high exposure and relatively higher LGDs, although no banks would breach minimum thresholds under the scenarios considered. - Note: There are currently no debt securities recorded under the amortized cost category; therefore all securities are marked to market. ### Recommendations from the FSAP - Differentiate inflation-indexed and non-indexed lending and funding in regular risk monitoring and stress testing: - Capture differing risk profiles via credit, interest rate, and market risk channels. - Credit risk: differentiate probability of default between indexed and non-indexed loans. - Interest rate models: capture different repricing schedules and interest dynamics for indexed vs non-indexed portfolios to gauge net interest margin impact. - Market risk: assess exposure to balance sheet revaluation risks due to inflation via net inflation-indexed position. - Improve data collection: non-performing loan ratio and probability of default separately for indexed and non-indexed loans. - CBI should periodically assess impact of interest rate shocks on bank profitability and solvency via net interest margin, credit losses, and market revaluation gains/losses on tradable and non-tradable securities, considering both short- and long-term horizons. - Enhance data quality and consistency of supervisory reporting: - Use automated validation rules to reject inconsistent banks’ reporting. ### Macro-financial feedback loops and methodology - Approach: - A structural VAR (SVAR) model links shocks to bank capitalization with paths for bank lending and macroeconomic variables. - Domestic variables: Iceland’s real GDP, CPI, nominal exchange rate (all in logs), unemployment rate, and policy interest rate. - External variables (exogenous): US real GDP, oil price (both in logs), and US policy rate. - Banking block: bank capital (accounting bank equity), bank outstanding loans to domestic households and non-financial corporates (logs and divided by CPI to express in real terms), and the spread between the lending rate and the policy rate. - Sample/estimation frequency: quarterly, over 1998Q1–2022Q2, using aggregate variables. - Identification: - Block exogeneity: bank capitalization impacts the macroeconomy only through bank lending. - Sign restrictions: a positive credit supply shock → positive impact on bank capital, bank loans, and real GDP; negative impact on unemployment and lending-policy rate spread. - Linking solvency stress-test to SVAR: - The change in bank capital between baseline and adverse scenarios from the solvency stress-test (maximum magnitude of 45 percent) was used to generate a credit supply shock in the SVAR. - Two-step conditional forecasting: (1) uncover structural credit supply shocks that generate the capital gap; (2) use estimated impulse responses to compute paths for endogenous variables. - Bridge equations estimated for variables not in SVAR (e.g., housing prices, stock market index) to generate conditional forecasts. ### Second-round scenario and macroeconomic impacts - Construction: - Second-round scenario = first-round adverse scenario + conditional paths from SVAR capturing macro-financial feedbacks. - Caveat: possible “double-counting” because first-round adverse scenario may already partially capture macro-financial feedbacks indirectly. - Key macro outcomes at trough (trough occurs in 2024): - Real GDP: 3.8 percent lower (additional cumulative reduction). - Unemployment: 1.6 percentage points higher (additional cumulative increase). - Nominal house prices: 9 percent lower (additional cumulative fall). - Nominal exchange rate: 11.4 percent depreciation (additional cumulative). - Change in inflation rate: negligible. - Change in policy rate: small, at a magnitude of 60bps or less in every year. - Deleveraging: - Second-round scenario implies deleveraging of lending portfolio equal to aggregate real loan growth from SVAR; deleveraging at trough is 14 percent (sizeable but much milder than GFC when real bank loans contracted by over 50% in the five years after 2008Q3 peak). ### Quantitative effects on bank capital and solvency (second round) - Aggregate CAR: - Second-round aggregate CAR is lower than first-round by 1.2 percentage points in 2024 and by 1.7 percentage points in 2027 (on average for the three banks). - Drivers: - Lower profitability is the dominant effect, driven mainly by higher credit risk from higher PDs (linked to higher unemployment). - Deleveraging lowers net interest income but reduces RWAs, which tends to increase capital ratios; net effect is that lower profitability outweighs RWA reduction. - Outcomes: - First round: all banks remain above hurdle rate throughout stress-testing period. - Second round: one of the three banks falls slightly below hurdle rate at the trough; aggregate capitalization remains above the hurdle rate. - Conclusion: aggregate solvency resilience confirmed, with vulnerabilities that could be addressed with macroprudential tools such as the CCyB. ### Calibration of the CCyB using macro-financial linkages (reverse-stress-testing) - Method: - Reverse-stress-test to find minimum CCyB rate such that, at the trough, all three banks remain above the CAR hurdle-rate throughout the stress-testing period. - Assumption: banks keep a constant buffer in excess of required capital levels; usable capital buffers include the CCoB, voluntary excess capital, and CCyB. - A higher CCyB increases required bank capital; the SVAR is used to forecast macro improvements conditional on this additional bank capital, producing a “third-round” scenario. - Quantitative outcome: - An additional 80bps of CCyB (relative to the 2 percent at the starting point) would be required for all banks to remain above the hurdle rate. - Resulting CCyB: 2.8 percent would keep all banks above the hurdle rate throughout the stress-testing horizon with the third-round scenario. - Caveats: - Calibration does not provide a full cost-benefit analysis; it informs calibration with respect to the particular stress scenario only. - Calibration could be over-estimated if macro-financial feedbacks were already partially captured in the first-round adverse scenario. - The improvement in macro outlook from higher bank capitalization is roughly offset by higher RWAs resulting from higher bank leverage in the model, making third-round and second-round based calibrations similar. *Source: IMF staff.* ### 58. Three distinct liquidity stress tests were conducted to assess bank capacity to ### 1islea2023003 - 58. Three distinct liquidity stress tests were conducted to assess bank capacity to ### A. Liquidity stress test design - Three distinct stress tests performed for the three D-SIBs in Iceland: LCR-based, cash-flow based, and NSFR stress test. - LCR-based stress test: - Measures ability to meet short-term liquidity needs in a 30-day horizon against the initial level of high-quality liquid assets (HQLA). - Cash-flow based stress test: - Leverages maturity profile over a 12-month horizon to assess maturity mismatches and availability of counterbalancing capacity. - NSFR stress test: - Used the NSFR limit of 100 percent (binding in 2021) to gauge structural long-term refinancing and funding risks. ### B. Data sources and thresholds - Main LCR, NSFR and cash-flow based stress tests based on COREP reports as of Oct-2022. - Comparison of liquidity position between 2019 and 2021 performed to assess buildup of liquidity buffers since the pandemic. - Data on resident and non-resident deposits obtained from CBI supervisory templates to assess cross-border funding reliance. - Thresholds used: - LCR and NSFR based stress tests used a 100 percent threshold (minimum regulatory requirement as of Oct-2022). - Current LCR regulatory minimums for individual significant currencies (as active in Iceland): 50 percent for Krona, 80 percent for Euro, and no LCR limit set for the US Dollar. For this exercise, Basel 100 percent LCR limit was imposed for total currency and significant individual currencies. - Cash-flow based stress test used counterbalancing capacity as the threshold; negative amounts indicate bank failure in the test. ### C. Liquid assets and funding structure — initial positions and evolution - Post-pandemic effects: - Public liquidity supports (central bank liquidity facility, low-cost long-term financing) and liquidity support to households/corporates boosted consumer deposits and bank HQLAs. - Freshly injected liquidity placed with the central bank as HQLAs, notably boosting liquidity indicators. - Recently, strong lending growth, dividend payments and share buybacks led to moderate contraction of bank liquidity positions, but still slightly above pre-pandemic level. - Contractual outflow increased due to maturing of existing debt issuance. - Banks are issuing new bonds to refinance existing debt and to fulfill ongoing MREL requirements; considering term deposits with shorter duration to attract funding. - Key metrics: - Loan-to-deposit ratio remains high at 150 percent. - LCR historical volatility by currency between 2019 and 2022Q3: - US dollar LCR fluctuates between 200 and 400 percent. - Euro LCR reached near 700 percent in 2022Q3, relative to end-2021 level at around 400 percent. - Funding and maturities: - 44 percent of funding placed within the overnight bucket. - Over 85 percent of cash inflows (mostly maturing loans) materialize beyond the first three months. - Most unsecured and covered bonds become due beyond 3 months. - Implication: maturity mismatch with frontloaded outflows and backloaded inflows raises vulnerability to sustained liquidity stress; need continued monitoring of maturity structures. ### D. LCR-based stress test scenarios and parameters (overview) - Six scenarios applied to three D-SIBs: - S1: Standard LCR (regulatory parameters per CRR). - S2: Retail stress — higher run-off rates for retail claims; banks can use liquid assets with no additional market-value decline (haircut follows CRR). - S3: Wholesale stress — higher run-off rates for wholesale claims; same haircut treatment as S2. - S4: Retail + wholesale (combo) — stressed run-off rates for both retail and wholesale (whichever is higher) and banks face haircuts when liquidating assets; haircuts draw on solvency stress test market-value declines and ECB valuation haircuts for repo to central bank. - S5: S4 + mild outflows from pension and foreign funding — additional outflows from deposit and maturing debt securities; no additional pension deposit outflow beyond standard LCR parameters; additional bond maturity outflows between cut-off (2022Q3) and current date included. - S6: S4 + severe outflows from pension and foreign funding — adds larger pension and foreign funding shocks, simulating severe non-resident deposit outflows and potential early redemptions. - Selected parameter examples from Table 2 (scenario-specific run-off / haircut calibrations preserved as described in source): - Stable retail deposits: S1 5%, S2 10%, S3 5%, S4 10%, S5 10%, S6 10%. - Other retail deposits: S1 10%, S2 20%, S3 10%, S4 20%, S5 20%, S6 20%. - Operational deposits: S1 5-25%, S2 5-25%, S3 15-35%, S4 15-35%, S5 15-35%, S6 15-35%. - Non-operational deposits other than financial institutions: S1 20-40%, S2 20-40%, S3 30-50%, S4 30-50%, S5 30-50%, S6 30-50%. - Non-operational deposits financial institutions: 100% across S1–S6. - Committed facilities to retail customers: S1 5%, S2 10-15%, S3 5-10%, S4 10-15%, S5 10-15%, S6 10-15%. - Committed facilities to corporate customers: S1 10-30%, S2 10-40%, S3 20-50%, S4 20-50%, S5 20-50%, S6 20-50%. - Pension funding (other than non-operational deposit): S1–S4 0%; S5 10%; S6 15%. - Foreign funding (other than non-operational deposit): S1–S4 0%; S5 15%; S6 25%. - HQLA haircuts (level 1, level 1 covered bonds, level 2A, level 2B) in S4–S6 include negative adjustments (e.g., level 1 covered bonds -20/-3%, level 2A -15/-5%, level 2B -25/-5%) informed by market-value declines and ECB valuation haircuts. ### E. LCR-based stress test results and implications - Aggregate LCR impact: - Starting point: 200 percent as of 2022Q3. - Under the most severe combo scenario (retail + wholesale) aggregate LCR fell to 122 percent; one bank’s LCR marginally below the minimum threshold among the three D-SIBs. - Adding additional outflows from pension and foreign funding reduced aggregate LCR to around 76 percent and caused one more bank to breach the minimum threshold. - Currency-level findings: - Significant-currency regulatory minima in Iceland: 50 percent for Icelandic Krona, 80 percent for Euro; no current regulatory LCR for US dollar (exercise assumed 100 percent for comparative purposes). - Under existing regulatory minimums one bank breaches the threshold in U.S. dollars. - All D-SIBs meet minimum threshold in Euro and Icelandic Krona under existing regulatory minima. - If a homogenous 100 percent threshold were assumed for all three significant currencies: - One bank breaches the minimum LCR threshold in Euro and one breaches in U.S. dollars. - None of the D-SIBs would be able to meet minimum threshold in Icelandic Krona under the most severe scenario. - Policy implication highlighted: - Need to develop approaches to continuously monitor funding risks from nonbank financial institutions (including pension funds) and foreign investors. - Consideration of separate monitoring and potential establishment of LCR requirements at 100 percent for individual significant currencies. ### F. Cash-flow-based stress test results and implications - Methodology: - Net liquidity position = cumulated net funding gap (inflows minus outflows across maturity buckets) minus cumulated counterbalancing capacity (liquid assets across buckets). - If net liquidity position becomes negative after using counterbalancing capacity, a liquidity shortfall is recognized. - Maturity profile and shocks: - About 50 percent of total outflows projected to take place in less than 30 days; open maturity bucket holds about 44 percent of total outflows. - Retail and corporate deposits (operational and non-operational) are main contributors to short-term funding. - Outflows from committed facilities considered immaterial at only 3 percent of total contractual outflows. - Bank-issued debt securities (secured and unsecured): over 90 percent maturing beyond one year. - Inflows: 80 percent of total inflows concentrated in loans and advances; of these, 90 percent have maturity longer than one year. - Calibration assumptions: - Higher run-off rates for wholesale than retail funding. - Secured funding sources have lower run-off rates than unsecured. - Non-resident deposits placed in components subject to higher outflow rates. - Inflow parameters are 100 percent of contractual inflows except loans to retail and corporate customers (0 percent) to reflect debt moratoria and assumption that banks cannot deleverage. - Findings: - Reliance on counterbalancing capacity for short-term shocks; potential liquidity shortfalls over longer term due to backloaded inflows. - Large lending exposure contributes to low asset encumbrance ratio: - Aggregated asset encumbrance ratio around 15 percent. - Low share of central bank eligible (CB-eligible) assets out of total unencumbered assets due to low share of debt securities. - Policy recommendation: - Banks could benefit from building up CB-eligible liquid assets to access additional funding sources (e.g., CBI liquidity facility or repo market) and solidify liquidity positions. *Source: IMF staff.* ### 72. The cashflow liquidity stress test runs a set of embedded scenarios of increasing ### The cashflow liquidity and NSFR-based liquidity stress tests (Iceland) ### Cashflow liquidity stress test: scenarios and methodology - Horizons: 5-days, 4-week, 3 months, and 12-month horizons. - Stress scenarios: three scenarios with increasing severity — mild market stress, medium market stress and severe market stress — applied to all banks. - Counterbalancing capacity (CBC) approaches: - Full CBC: fully endogenous liquidity supply by the central bank as long as banks have unencumbered eligible collateral. - Full CBC with market haircuts: a full CBC is assumed, but market-specific haircuts and bank-specific market price effects are imposed on elements of the CBC. - Calibration: - Draws on past stress episodes using historical time series, previous FSAP stress parameters and regional experience in liquidity outflows in the Euro Area and Nordic region. - Takes into consideration structural improvements in the Icelandic banking sector since the GFC. - Market haircuts to CBCs draw from the outcome of market revaluation shock under the solvency stress test and are informed by the ECB valuation haircut when banks need to repo the liquid assets to the central bank. - Cash flow analysis spans contractual horizons up to 12 months (longer than the LCR 30-day horizon), covering both covered and uncovered bond maturities. ### Cashflow stress test parameters (selection of run-off and haircut assumptions) - Run-off Factors (Range in Percent across Mild, Medium and Severe Scenarios): - Unsecured bonds: 40-100% - Regulated covered bonds: 25-70% - Securitisations and others: 100% - Repos across all asset classes: 100% - Stable retail deposits: 2-10% - Other retail deposits: 5-20% - Operational deposits: 5-30% - Non-operational corporate deposits & other: 20-100% - Derivatives: 100% - Committed facilities: 10-100% - Outflows due to downgrade triggers: 0-100% - Reverse repos across all asset classes: 100% - Loan inflows from retail and corporates: 0% - Loan inflows from central banks: 100% - Loan inflows from banks and NBFIs: 30-100% - Loan inflows from others: 0-30% - Haircut Based on Market Price (levels expressed as percentages applied to asset categories): - Level 1 assets: 95% - Level 1 covered bonds: 90% - Level 2A assets: 85% - Level 2B assets: 50-75% - Other tradable assets: 50% - Non tradable assets: 50% ### Cashflow stress test: key findings (system-level and bank-level) - System-level: - Cash shortfall over a 12-month horizon under the most severe scenario amounts to 283 billion, roughly 6 percent of total asset. - Over 30-day horizon, banks on aggregate can broadly withstand liquidity outflows supported by existing counterbalancing capacities. - Bank-level heterogeneity: - Two banks would already experience liquidity shortfalls even within 30 days under the most severe scenario with a total cash shortfall at around 2.7 percent of total assets. - Bank-specific shortfalls over the 12-month horizon are more pronounced than system level shortfalls, at 348 billion, equivalent to 7.3 percent of total assets. - Interpretation: maturity mismatch with frontloaded inflows and backloaded outflows (most unsecured and covered bonds due beyond 3 months) weakens liquidity positions beyond 30 days. ### Currency-specific cashflow stress test results - Separate cashflow-based stress tests conducted for Icelandic Krona, Euro and U.S. dollar using same assumptions as the total currencies test. - Shortfalls (bank-specific) over 30-days (12-months): - Euro: 0.03 (0.4) billion representing 0.08 (1.3) percent of total banking assets. - U.S. dollar: 0.16 (0.2) billion representing 0.5 (0.6) percent of total banking assets. - Icelandic Krona: 108 (234) billion representing 2.3 (4.9) percent of total banking assets. - System-level shortfalls over 30-day horizon: - Euro: no liquidity shortfalls identified. - Icelandic Krona: 5.1 billion representing 0.1 percent of total assets. - US dollar: 0.2 billion representing 0.5 percent of total assets. - System-level shortfalls over 12 months: - Euro: 0.4 billion representing 1.2 percent of total banking sector assets. - US dollar: 0.2 billion representing 0.6 percent of total banking sector assets. - Icelandic Krona: 181 billion representing 3.8 percent of total banking sector assets. - Observations: banks’ liquidity positions are weaker in individual currencies, such as Icelandic Krona and US dollar, especially over the short-term. Bank-level results show individual banks experiencing shortfalls in specific currencies even when system-level shortfalls appear manageable. ### NSFR-based liquidity stress test: context and starting metrics - Aggregate NSFR as of 2022Q3: - Aggregate NSFR of sample banks: 123 percent (above minimum requirement of 100 percent, with no single bank below the threshold). - NSFR by currency as of 2022Q3: - Euro: around 264 percent - Total currency: 123 percent - Icelandic Krona: 111 percent - US dollar: around 93 percent - Funding composition differences: - Long-term Euro funding mostly came from other banks. - US dollar funding originated mostly from non-financial investors. - Additional considerations: - Banks face binding constraint from MREL requirement (issuance of sufficient long-term subordinated debt). - High share of long-term loans and limited HQLA assets lead to relatively higher needs of available stable funding (ASF). ### NSFR stress test methodology - Volume-based technique simulating migration from long-term to short-term funding. - Migration assumptions: - 50 percent of the funding within six- to twelve-month bucket would flow to less than six-month bucket. - 35 percent of the funding with more than one year maturity would migrate to the six- to twelve-month bucket. - 15 percent flow rate from over one-year bucket to the less than six-month bucket. - Applicable required stable funding (RSF) and ASF factors are maintained under the stress scenario. - Equity instruments were not stressed (assumed banks will not conduct stock transactions such as new issuances or repurchases). ### NSFR stress test results - Aggregate outcome: - Post-shock total currency NSFR saw no bank falling below the 100 percent threshold. - On aggregate, larger decline of NSFR in Euro than in total currencies and Icelandic Krona, though both remain above 100 percent. - NSFR in US dollar continues trailing other currencies at below 100 percent. - Bank-level: - One bank failed below the threshold for Euro. - One bank failed below the threshold for US dollar. - Two banks failed below the threshold for Icelandic Krona. ### Recommendations (FSAP) - CBI should enhance monitoring of LCR by currencies. - Introduce separate analysis on individual significant currencies, such as Icelandic Krona, Euro and US dollars, leveraging existing LCR and cashflow stress test frameworks. - Address outlier banks with weak liquidity positions through Pillar 2 and supervisory actions to ensure banks have adequate FX liquidity buffer in each currency to withstand corresponding outflows over both short term and long term. ### Bank interconnectedness and contagion testing (method and indices) - Contagion model: - Uses network model developed by Espinosa-Vega and Sole (2010). - Simulates cascading failures due to credit and funding shocks; a bank defaults when CET1 ratio drops below 4.5 percent. - Parameters include λ (loss ratio on exposure claims during a credit shock), ρ (fraction of funding lost in a funding shock), and δ (asset price discount when assets are sold to replace lost funding). - Output indices: - Index of contagion (Cont_i): 100 * (1/(N-1)) * Σ_{j=1,j≠i}^N (L_j i / K_j), where L_j i is the total capital loss of bank j due to bank i’s default and K_j denotes capital of bank j. - Index of vulnerability (Vuln_i): 100 * (1/(N-1)) * Σ_{j=1,j≠i}^N (L_i j / K_i), where L_i j is the total capital loss of bank i due to the default of bank j and K_i denotes capital of bank i. *Source: IMF staff.* ### 87. Iceland’s domestic banking system is concentrated and interconnected with Non-bank ### 1islea2023003 - 87. Iceland’s domestic banking system is concentrated and interconnected with Non-bank ### Domestic banking concentration and interbank contagion - The banking sector comprises four commercial banks and five saving banks, taking up 146 percent of the GDP. - The system is dominated by three commercial banks, accounting for 95 percent of total banking assets; the interbank exposure matrix thus focuses on the three largest commercial banks. - Interbank exposure data cover each bank’s interbank assets and liabilities vis-à-vis each other as of 2022Q2. - Key findings from the domestic interbank contagion analysis: - Contagion risks through credit and funding channels are very limited. - Domestic interbank exposures are small, especially relative to the capitalization of banks. - For the three banks, gross domestic exposures are much smaller than their regulatory capital; tests show resilience under a credit shock and under concurrent credit and funding shocks. - No single failure of a domestic bank would trigger the failure of other banks in the system. - None of the three banks are found to be undercapitalized after shock. (Undercapitalization is when the bank’s CET1 level drops below 4.5 percent.) - Bank 1 is the most contagious: its default would bring the most severe percentage of capital loss to other banks. - Bank 2 is the most vulnerable: the capital loss rate of bank 2 due to the default of all other banks is the most significant. ### Bank–NBFI interlinkages and contagion - Network composition: three systemic banks plus the largest 15 pension funds and the largest 15 investment funds. - Structure of exposures: - Pension funds are the primary creditors, having 80 percent of asset exposures from banks and investment banks. - Banks are the primary borrowers, holding 94 percent of the total liabilities in the Bank–NBFI system. - Contagion and vulnerability results: - Banks are the most contagious in the case of a credit shock, causing the most severe capital loss to other institutions. - Contagion indices of pensions and investment banks remained limited. - One pension fund and several investment banks are found to be more vulnerable to credit shock than others, albeit the overall vulnerability level for all institutions remained under 2.5 percent. - Banks are the most vulnerable to funding shocks, given their role as the primary borrower in the network. ### Cross-border interbank contagion - Capital mobility and context: - By 2022 Iceland’s degree of capital mobility levels was at least as high as in the early 2000s, or higher. - All capital flow management measures introduced during the global financial crisis have been removed; the reserve requirement rate on selected capital inflows adopted in 2016 has been set to zero but remains a possible tool in the legal framework. - Net foreign asset positions (by end-2022): - Icelandic banks have an aggregate negative net foreign asset position of about 13 percent of GDP (about 60 percent of capital). - The sizable international reserves of the central bank turn the net foreign position of the banking system positive. - Significant savings of pension funds and other institutional investors abroad further increase the positive net foreign asset position of the financial system. - Exchange rate and FX risk management: - Banks’ exchange rate risk is limited because banks are subject to conservative limits in the net open foreign exchange positions (in percent of capital). - Banks hold foreign currency denominated assets (deposits abroad, loans to nonresident corporations, direct loans to Icelandic firms that generate foreign exchange, derivative coverage). - Banks are subject to regulations on the liquidity of their foreign currency portfolios and these positions’ liquidity under stress is tested in the systemic liquidity analysis. - Cross-border liabilities and policy backstops: - Most bank cross-border liabilities arise from the issue of medium-term foreign currency denominated bonds; this protects against immediate withdrawals but exposes banks to refinancing risks when amortizations are due. - In stressful situations, the central bank could provide FX support against collateral including FX swaps or intervene in the foreign exchange market; alternatively, banks may issue FX debt instruments, such as FX denominated covered bonds, that pension funds could purchase if price and recovery prospects are right. - Cross-border contagion stress analysis (data and model): - Uses supervisory data of Icelandic banks' total exposures to foreign counterparties and foreign banks' consolidated claims on Iceland from the BIS; data reporting date is 2022Q1. - Espinosa-Vega and Sole (2010) model parameters: lambda is set to 70 percent, delta is 50 percent, and rho is 35 percent. - Cross-border exposures are geographically clustered: over two-thirds of Icelandic banks' foreign exposures come from Europe; around 30 percent of exposures are vis-à-vis the U.S. and Canada. - Results: - Icelandic banks are not a source of contagion risk to other major economies but can be negatively affected due to inward spillovers from other financial centers. - The Icelandic banking sector remains resilient after simulated credit shocks and capital level is above the regulatory minimum. - The inward spillover from the U.S. is the most significant among the sample countries analyzed. ### Market-based interconnectedness analysis - Method and data: - Applied Diebold and Yilmaz (2014) measures to capture financial market spillovers using daily stock price indices and sovereign CDS returns covering 14 and 11 countries respectively; data from Bloomberg and Haver; reporting date between 2009/01/06 and 2022/11/21. - Returns computed as a two-day log difference to control for global trading hour differences. - Key findings: - Global financial conditions have a strong effect on Icelandic financial conditions. - The Icelandic stock market is highly interconnected with foreign markets; Iceland is a net receiver of stock return spillovers. - Among closely linked countries, France, the Netherlands, and Germany are the primary equity return volatility contributors to Iceland; except for the United States, net contributors are clustered in Europe. - Iceland is a net receiver of sovereign CDS return spillovers; among the sample, Iceland is the primary net receiver of sovereign CDS return spillover. - Italy appears as a primary source of net return connectedness to all other countries in the sample, followed by France and Belgium. ### Systemic liquidity analysis — overview and sectoral linkages - Motivation: - As a small open economy with a concentrated banking system, Iceland is highly exposed to adverse systemic liquidity events from domestic and global factors (foreign investor exits, FX deposit withdrawals, outflows from bond maturities, sales of domestic sovereign securities), potentially causing asset price spirals and reputational contagion across the three similar D-SIBs. - Objectives of the systemic FX liquidity stress test: 1. Identify key sectors providing FX liquidity. 2. Identify key transmission channels of an FX liquidity shock (direct withdrawals and currency conversion into FX from local currency). 3. Assess adequacy of FX liquidity buffer in the banking sector to withstand FX outflows (from pension funds, investment funds, foreign investors). 4. Assess the capacity of the CBI to provide FX liquidity support to meet market FX demand in stress, especially to commercial banks. - Data and outputs: - Uses bilateral sectoral claim data at the aggregated level as of end-2021; covers central bank, commercial banks, pension and other NBFIs (including investment funds), corporates and households, and foreign investors or borrowers. - Assumes a combination of FX liquidity shocks with increasing severity: nonresident deposit outflows, household and nonfinancial corporate withdrawals, asset reallocation of pension and other investment funds, and potential outflows from maturities of international bonds. - Main outputs: ending FX liquidity balance of the banking sector and level of international reserves pre- and post-shock. - Stylized balance sheet findings: - Asset managers (pension and other NBFIs) tend to hold more FX assets than FX liabilities and are net FX lenders; most FX assets are highly liquid and denominated in bank FX deposits or marketable debt securities. - Commercial banks have a balanced FX position overall (issuing FX deposits and bonds to nonresidents; holding FX loans to the private sector and deposits to other banks; investing in foreign sovereign securities). - Excluding less-liquid assets (foreign currency loans), commercial banks reveal certain FX imbalances and appear to have limited FX liquidity buffer to meet large FX withdrawals under rapid funding stress. - Central bank reserves and related magnitudes (as of end-2021): - Gross international reserves stand at around 6.5 billion US dollar. - Out of total FX borrowings, roughly 67 percent are through the issuance of government bonds denominated in foreign currencies and placed at the CBI to boost gross international reserves; the rest comes from direct foreign liabilities of the CBI. - Net international reserves, when netting out both forms of FX borrowings, amount to roughly 3.7 billion US dollar. - Role of pension funds: - Pension funds’ assets abroad can provide stability in certain capital outflow scenarios by purchasing assets left behind by nonresidents, thereby alleviating pressure in both foreign exchange and domestic bond markets. - If pension funds exceed foreign exchange investment limits, exchanging assets abroad for domestic assets may not be beneficial for pension funds and could affect portfolio returns. - In deep crises with long-term economic scars, pension funds could temporarily repatriate overseas assets in the interest of pensioners if returns and market exchange rates are appropriate. *Source: Supervisory data, BIS consolidated banking statistics, Bloomberg, Haver, and IMF staff calculations as presented in the document.* ### 105. The stress test simulates various liquidity outflow scenarios with increasing severity ### The stress test simulates various liquidity outflow scenarios with increasing severity ### Stress test design and transmission channels - The exercise simulates a joint materialization of outflows from: - domestic household and corporate deposit, - nonresident deposits, - international bond maturity, - additional funding shock from pension and other NBFIs stemming from reallocation of assets from onshore to offshore. - Combined FX liquidity shocks are imposed with increasing severity to simulate high co-movement under stress. - Shocks are benchmarked against either the existing LCR parameters over a 30-day horizon for household and corporate deposit outflows, or historical stress episodes observed in Iceland such as runoff rate for nonresident deposits. - A 30 percent currency depreciation was assumed across all the adverse scenarios. - Local currency conversion is allowed in some scenarios to capture conversion of domestic assets into foreign currency, amplifying FX outflows. ### Shock parameters (ranges preserved exactly as in the source) - Shock 1 HH FX deposit shock: [10,20] based on LCR; Local Currency Conversion Y [5,10] - Shock 2 NFC FX deposit shock: [20,40] based on LCR; Local Currency Conversion Y [10,20] - Shock 3 Nonresident deposit shock: [20,80] based on LCR and historical stress episodes; Local Currency Conversion N/A - Shock 4 International bond outflows: [15,30]; Local Currency Conversion N/A - Shock 5 Funding shock from pension funds and other NBFIs due to share redemption and asset reallocation: [20,40]; Local Currency Conversion Y [10,20] - Shock 6 FX liquid asset haircut: [2,20] based on LCR and solvency stress test, lower haircut for global safe assets, higher haircut for other domestic FX assets; Local Currency Conversion N/A ### Behavioral and operational assumptions - Banks are assumed to exhaust existing FX reserves before liquidating other FX assets. - Other sectors use local FX bank deposit first before using other FX liquid assets to meet FX liquidity outflows. - Local currency conversion by investors (household, corporate, nonresident investors, pension funds and other NBFIs) is incorporated in scenarios to augment capital flight. ### Results: banking sector FX liquidity and scenarios - Baseline scenario (only normal inflows and outflows): banks generally have sufficient FX liquidity buffer to meet regular financial and operational needs without FX liquidity support from the CBI. - Mild scenario (mild deposit and bond funding outflows without local currency conversion): banks show a comfortable level of FX liquidity buffer post shock. - Severe scenario (larger shocks for all types of outflows; local currency conversion allowed): banks could face sizable FX liquidity gaps. - Maximum FX liquidity gap under severe scenario: 3.4 billion US dollars, about half of the gross international reserves. - The gap would be larger should pension funds or other NBFIs decide to move some domestic assets offshore. - Finding warrants developing approaches to continuously monitor banks’ liquidity risks associated with funding shocks from nonbank financial institutions (including pension funds) and foreign investors. ### International reserves and central bank backstop capacity - Both gross international reserves and net international reserves remain positive under the most severe scenarios. - There is currently no FX reserve requirement imposed on Icelandic banks by the CBI; required reserve, excess reserve and total reserve are set to 0 in the analysis. - FX buffers from the CBI may be channeled to banks via direct central bank liquidity support or swap agreements with other European banks. - Of total FX government bonds outstanding: - 500 million Euro is scheduled to mature in 2024. - 500 million Euro will be paid off in 2026. - 750 million Euro will be paid off in 2027. - Assuming an additional 500 million Euro outflow from the CBI to meet the 2024 maturity, gross international reserves would decrease by same amount, without impacting the net international reserves. - The net international reserves post shock becomes significantly lower, due to initial large netting effect from the local government FX deposit (obtained via the issuance of FX government bond) placed at the CBI. ### Single-factor sensitivity analysis: ranges and tipping points - Outflow rates considered independently: - International bond outflows: 20 to 60 percent. - Non-resident FX deposit outflows: 40 to 100 percent. - FX funding outflows from pension fund and other NBFIs: 20 to 40 percent, along with local currency conversion from 10 to 30 percent. - FX deposits from domestic households: 10 to 20 percent, along with local currency conversion from 5 to 15 percent. - FX deposits from domestic corporates: 20 to 40 percent, along with local currency conversion from 10 to 50 percent. - Key sensitivity findings: - Only a 60 percent international bond outflow and a 100 percent non-resident deposit outflow (upper bound shocks) would lead to marginal FX liquidity gaps in the banking system. - Banks can withstand stand-alone shocks from: - households up to 20 percent (plus LC conversion up to 15 percent), - corporates up to 30 percent (plus LC conversion up to 30 percent), - NBFIs up to 30 percent (plus LC conversion up to 20 percent). - If LC conversion for NFC goes above 50 percent, the banking system would experience FX liquidity pressure leading to a small FX liquidity gap of 154 million US dollars, equivalent to a 2.6 percent reduction of the gross international reserves. - If LC conversion for NBFI goes above 30 percent, the banking system would experience a small FX liquidity gap of 320 million US dollars, equivalent to a 5.4 percent reduction of the gross international reserves. ### Pension fund market: scope, size, and asset composition (exact figures preserved) - Pension fund market size and structure: - Assets of almost ISK 7 trillion at end-2022, or 176 percent of GDP. - Pillar II schemes manage ISK 5.9 trillion. - Pillar III providers manage ISK 1.0 trillion. - Pillar II offered by 21 pension funds. - Pillar II four largest providers hold a cumulated market share of 61 percent. - Pillar III: two large providers account for almost half of the market. - Asset composition and exposures: - Directly held stocks account for 19 percent of Pillar II and 7 percent for Pillar III assets. - Around one third of Pillar II assets is invested in inflation-linked fixed-income assets, including mortgages which account for 10 percent of assets. - Combined exposure of pension funds to domestic banks, corporates and the sovereign increased from ISK 250 billion in 2017 to ISK 670 billion in 2021, or from 6 to 10 percent of pension fund assets—this corresponds to 14 percent of banks’ financial liabilities. - In Pillar III these exposures are mostly in the form of deposits, while in Pillar II they are mostly bonds and stocks. - As of end-2022, 37 percent of Pillar II assets were denominated in foreign exchange (24 percent in Pillar III), mostly in US dollar. - Resulting currency risks are typically not hedged. - Historical returns and recent performance: - From 1995 to 2021, Pillar II pension funds yielded 4.3 percent on average in real terms. - 2022 real return for the median fund: -12 percent. - Funding ratios of several DA schemes have dropped below 100 percent after 2022, but the large majority remained within tolerance limits not requiring adjustments to pension benefits. ### Policy recommendations (preserved wording and emphasis) - The FSAP recommends CBI to develop stress testing approaches to assess funding risks from nonbank financial institutions (including pension funds) and foreign investors. - Regularly assess concentration risks of funding between sectors. - Conduct systemic liquidity stress analysis to simulate banks’ loss of funding from NBFIs (including pension funds) and foreign investors, should they change their investment behavior by re-allocating assets cross border. - Ensure the capacity of the CBI to provide the needed FX liquidity support to the financial system in times of stress. *Source: IMF staff, extracted from the Iceland FSAP chapter on systemic liquidity and pension fund stress testing.* ### 117. For most parts of the analysis, the sample included the seven largest pension funds, ### Pension fund analysis (Iceland) ### Sample and fund characteristics - Sample covered the seven largest pension funds, achieving a market coverage of 77 percent in terms of Pillar II DA assets. - Size dispersion within the sample: - Based on assets, the largest pension fund is about nine times larger than the smallest. - Based on active members, the difference between the largest and the smallest entity amounts to a factor of four. - Member and contribution heterogeneity: - Funds differ in average age of members and average annual contributions; some funds target primarily blue-collar or white-collar employees. - Investment composition: - Differences in asset composition are much less pronounced across funds. - Funds with younger membership typically invest more in shares. - Share of foreign-denominated assets ranges from 33 to 42 percent. ### Scenario and methodology - Adverse scenario: - Derived from the adverse macrofinancial (stagflation) scenario used for the banking sector ST, including intensifying spillovers from Russia’s war in Ukraine, an abrupt global economic downturn, and a sudden correction in Icelandic asset markets. - Key market and credit risk variables: interest rates, stock prices, the external value of the Krona, and the inflation rate. - Projection approach: - Analysis focuses on the difference at retirement age between the baseline and the adverse scenario. - Representative members projected: those with 10 or 30 years to retirement. - Market risk stresses applied in each of the first three years of the projection horizon, 2023–2025; afterwards annual investment returns align with the baseline. - After t+3y, baseline returns are based on bootstrapping of historic returns for individual asset classes and each pension fund’s asset allocation as of the reference date. - Assumes retirement funds maintain asset allocation over the full projection horizon and re-balance annually. - Contributions during 2023–2025 are higher in the adverse scenario than in the baseline due to higher inflation. - Contributions after t+3y are assumed to increase at a constant rate of 4.0 percent per year. - Representative simulated member profiles: - 10 years prior to retirement; ISK 25m accrued pension savings; ISK 1.2m annual contributions. - 30 years prior to retirement; ISK 10m accrued pension savings; ISK 1m annual contributions. ### Impact on future pension values (adverse vs baseline) - Asset value path for the median pension fund: - Decline of 13 percent in 2023. - Further decline of 3 percent in 2024. - Partial recovery of +10 percent in 2025. - Most valuation impact stems from lower stock prices; in the first year ISK depreciation increases the value of foreign-denominated investments, partly offsetting losses. - Effect on members close to retirement (10 years prior): - Future pension values decline by between 8 and 15 percent across funds. - Median fund decline: 13 percent. - Effect on younger members (30 years prior): - Median fund decline: less than 1 percent. - Range across funds: minus 2 percent to plus 0.5 percent. - Note: In some cases future pension values in the adverse scenario can be higher than the baseline because higher inflation raises contributions, causing an upward level shift persisting until retirement age. ### Sensitivity analyses - Discount rate shock: - Liability duration of funds: on average around 30 years. - Lowering the discount rate from 3.5 percent to 3.0 percent would: - Increase the value of liabilities by 15 percent (very uniform across the sample). - Deteriorate the actuarial position by almost 10 percentage points overall. - Future position would decline by -11 percentage points. - Accrued position would decline by -8 percentage points. - Mortality (longevity) shock: - Many funds implemented new mortality tables in 2021/22 with dynamic longevity adjustments. - A decline of mortality rates by 10 percent across all age cohorts would: - Increase liabilities by around 2 percent. - Reduce the actuarial position by 2 percentage points. - All funds in the sample showed very similar sensitivities to changes in the discount rate and the mortality decrease. ### Liquidity risks - Pillar II: - Liquidity risks contained: inflows and outflows are predictable (no withdrawals; outflows tied to retirement, death, disability). - Sector is still accumulating funds; contributions exceed pension payments. - As of end-2022, large pension funds held no derivatives or off-balance instruments from which margin calls could arise. - Pillar III: - Cash flows can be more volatile due to member behavior (transfers between funds, mortgage repayments deducted from contributions, extraordinary withdrawals). - Extraordinary withdrawals allowed during the Financial Crisis and the Covid-19 pandemic: - Total withdrawals in 2020: ISK 23 billion. - Total withdrawals in 2021: ISK 13 billion. - Quarterly outflows of those funds that experienced an outflow have not exceeded ISK 100 million in any given quarter. - Net cashflows have remained positive in all quarters for almost every pension fund in the sample. - Structural changes in flows: - Prior to 2020, inflows to investment funds and the ratio of Pillar III premiums to disbursements tended to be higher; both declined after 2019. - Gross savings rate context: around 26 percent between 2017 and 2019, and 21 percent in 2020 and 2021. ### Further asset-side vulnerabilities - Mortgage lending: - Pension funds hold a share of around 22.7 percent in the mortgage lending market as of end-2022. - Indexed mortgage loan volumes stagnated between ISK 6 and 8bn per quarter since early 2021. - Non-indexed mortgage loan volumes picked up significantly in 2022, especially loans with temporarily fixed interest rates. - Mean LTV ratios of newly issued loans have been fluctuating around 50 percent recently. - Mean DSTI ratios increased to slightly above 20 percent for most pension funds as of mid-2022. - Risk: rising interest rates and stretched household finances could increase default probabilities and losses. - Banking sector exposures: - All large pension funds have concentrated exposures to the three large domestic banks. - For each of the two largest banks, exposure averages more than 3 percent of pension fund assets. - Exposure to the whole domestic banking sector amounts to slightly more than 10 percent. - Banking exposures composition: 45 percent (covered bonds), 38 percent (shares), and 14 percent (deposits). - Single-name concentrations remain below prudential investment limits in the Pension Fund Act, but sectoral concentration warrants closer monitoring. - Pension funds have been a steady source of funding for domestic banks; deposits experienced withdrawals up to ISK 10 billion in several quarters. - Investment behavior: - Pillar II funds are generally long-term investors and have not changed strategic asset allocation frequently. - Since 2018, Pillar II has shown almost no change in annual net equity investments. - Between 2019 and 2021 bond exposure was reduced; trend may have reversed as higher interest rates made bonds more attractive. - Pillar III equity investments are more volatile; Pillar III has been a net buyer of bonds since 2016. ### Recommendations - Data and supervisory reporting: - CBI should continue improving supervisory reporting quality and quality assurance. - Use of automated validation rules is recommended to reject inconsistent pension funds’ reporting. - Monitoring in a high inflation / high interest environment: - CBI should closely monitor effects on pension funds if higher inflation and interest rates persist. - Monitor funds’ investment behavior and counterparty default risk if corporate interest coverage deteriorates. - Monitor pension fund members’ behavior: propensity to switch Pillar III providers and transfers, and use of contributions to repay mortgage loans. - Liquidity analysis focus should be on smaller pension funds and those with already volatile Pillar III cash flows. - Mortgage lending oversight: - CBI should further analyze pension funds’ mortgage lending practices. - Investigate whether mortgage pricing reflects underlying risks, provisioning for expected losses, and pension funds’ risk management practices. *Source: IMF staff based on CBI data and company submissions.* ### 137. Stress test analysis imposed macroeconomic scenarios on the end-2020 firm-level data ### Stress test analysis imposed macroeconomic scenarios on the end-2020 firm-level data ### Stress test methodology and data - Main firm-level data source: ORBIS (Bureau van Dijk) reports about 24,000 non-financial companies for the 2020, the latest vintage available when the analysis was done. - The NFC sample represents about 70 percent of active firms reported by Statistics Iceland. - The sample is revised down significantly to remove outliers and eliminate dormant firms through the analysis period. - Baseline and adverse macro scenarios for the NFC analysis are in line with the bank stress testing and additionally use inputs from corporate bond spread, short- and long-term funding cost, and a financial condition index to generate results for the dynamic scenario-based stress tests under baseline and adverse scenarios. - Financial condition index (FCI) scenarios use historical data from the recently published FCI by the CBI; baseline and adverse scenarios for the FCI are constructed drawing mainly on the historical correlation between macro and financial indicators. ### Sensitivity analysis findings - For 2020 cross-section: - Median firm: interest coverage ratio (ICR) was already constrained, profits (ROA) were low, and the debt ratio was high, despite a reasonable cash coverage. - Interest expense shock: - A 30 percent change in interest expenses increases the share of firms at risk by 5 percent and debt at risk by 13 percent for the firms with interest coverage ratio falling below the threshold. - A hike in interest expenses also yields a surge in the share of firms with cash balance below zero and in the share of firms with equity below zero, and subsequently their related shares of debt at risk. ### Scenario analysis results (baseline and adverse) - Baseline scenario (2020 to 2022, with 2022 as the base year for the adverse scenario): - Share of debt for firms with an ICR<1 increases from 43 percent to about 70 percent within a year. - Share of debt for firms with cash<0 increases from 68 percent to 75 percent within a year. - Probability of default increases by about one percent. - Adverse scenario (large slowdown for two consecutive years, followed by gradual recovery): - ICRs deteriorate substantially. - Share of debt for firms with an ICR<1 rises to 96 percent in the first year, 26 percentage point higher than under the baseline. - Share of debt in firms with cash<0 surges 25 percentage point higher than under the baseline. - Probability of default peaks at 14.3 percent in the first year, about 3 percentage points higher than under the baseline, and remains high in subsequent years. ### Capital flow management measures (Appendix I) - Legal framework: - Explicitly considers possibility of adopting capital flow management measures to support financial stability and should abide by Iceland’s international obligations under OECD and EFTA agreements. - Capital outflow controls: - Reserved for emergency circumstances when large capital flows can endanger financial stability. - The central bank, with approval from the Ministry of Finance, can restrict or halt the flow of selected types of investments for up to 60 days without parliamentary approval. - Preemptive inflow measures: - The central bank, with approval of the Ministry of Finance, is allowed to set reserve requirements on selected inflows into krona-denominated assets (bonds, deposits, and unit shares in funds and other financial arrangements that invest in krona assets). - These reserve requirements were adopted in 2016; its rate originally was set at 40 percent and was gradually reduced and set to zero in 2019. - The design is similar to those adopted by Chile in 1991; studies of that experience showed they did not affect the total amount of capital inflows but did affect their composition. ### Risk Assessment Matrix (Appendix II) — selected risk entries and impacts - Intensification of regional conflict(s): - Relative Likelihood: High - Impact if Realized: Commodity price shocks, global slowdown, subdued tourism, tighter financial conditions, higher credit risk. - Abrupt global slowdown or recession: - Relative Likelihood: Medium (U.S.) / High (Europe) - Impact if Realized: Spillovers through trade and financial channels, downward pressure on commodity prices, knock-on effects from higher risk spreads and external financing costs, lower tourism earnings, rising unemployment, housing market correction. - Sudden correction in the domestic real estate market: - Relative Likelihood: Medium - Impact if Realized: Higher impairment charges for banks, defaults or delayed loan repayments by highly leveraged households, depressed domestic demand via reduced consumption, pressure on banks’ profits. - Systemic financial instability: - Relative Likelihood: Medium - Impact if Realized: Sharp correction in asset prices leading to valuation losses for banks, intensified credit risks from large interest rate swings, volatility in net interest margins potentially leading to further profit losses. ### Growth-at-Risk (GaR) analysis (Appendix III) - Framework: - GaR aggregates a large set of variables into seven main regressors (partitions) using principal component analysis (PCA): Housing market conditions, Stock market conditions, Money market conditions, Bond market conditions, FX market conditions, Main trading partners macro conditions, World financial conditions. - Data coverage started in 2003 at a quarterly frequency. - Example variables included: - Difference between Euribor 3m and T-bill Euro 3m; STOXX 50 Volatility Index; Household interest expense to disposable income; Total loans to non-financial corporate sector; USA unemployment rate; Term spread of EU (10-year yield EU - 1 year yield EU); VIX; Residential Real Estate Prices; CBI key policy rate; Real effective exchange rate; France unemployment rate; EU unemployment rate; Yield on 10-year inflation-indexed government bond; Long-term interest rate differential between Iceland and Germany. - Quantile regression approach: - Quantile regressions estimated for q in {5%, 25%, 50%, 75%, 90%} to map predictors to future GDP growth quantiles. - A skewed t distribution is fitted on the empirical conditional quantile function for each horizon to estimate tail risks. - Key GaR result (based on financial conditions at 2022:Q3): - Severely adverse outcome (5 percent left tail): GDP growth falls below 2.3 percent one-year ahead. - Severely adverse outcome (5 percent left tail): GDP growth falls below 9.1 percent in the two-year horizon. ### Macro-Financial linkages and credit supply shock (Appendix IV) - Regression of real bank loans on bank capital: - Regression specification uses log outstanding stock of loans to domestic HHs and NFCs divided by CPI as dependent variable and log banking system total equity and minority interests divided by CPI as bank equity regressor (with a lag). - Estimated coefficient interpretation: a 1 percent increase in bank capital leads to a 0.12-0.17 percent increase in bank real loans (the range 0.12-0.17 corresponds to different sets of controls). - Structural credit supply shock estimated with Bayesian SVAR: - Log bank real equity chosen as measure of bank capitalization. - Specification with 2 lags was chosen. - Identification based on block exogeneity assumptions and sign restrictions; sign restrictions imposed for quarters 0 and 1 after the shock (unemployment sign restriction imposed for quarters 0 through 8). *Source: ORBIS and IMF staff calculations; content drawn from the provided IMF chapter excerpt.* ### 3. Figure 1 displays the median impulse responses obtained from the Bayesian SVAR ### 3. Figure 1 displays the median impulse responses obtained from the Bayesian SVAR ### Bayesian SVAR — Credit Supply Shock Impulse Responses - The charts display the median impulse-response functions together with one-standard deviation bands obtained from the Bayesian SVAR. - Key empirical findings: - The positive impact of a credit supply shock on bank capital and lending is very persistent. - Confidence bands are quite wide, reflecting significant uncertainty in the estimation; this could be due to the limited number of observations and the large shocks suffered by the banks within the sample. - The negative impact on economic activity, as captured by GDP and unemployment, is more short-lived. - The IRFs for the nominal exchange rate, the CPI and the policy rate are not statistically different from zero. - The spread between the lending and policy rates goes down on impact and fades out after a few quarters. - Identification note: - Imposing the sign restriction for quarters 0 through 8 instead of 0 and 1 for unemployment yielded an IRF with a more sensible shape. - Variables in the SVAR (endogenous / exogenous as displayed): - Endogenous: ln(real equity), ln(real loans), ln(real GDP), Unemployment, ln(NEER), ln(CPI), Policy rate, Lending-Policy spread. - Exogenous: ln(US real GDP), ln(Oil price), US FFR. *Sources: IMF staff calculations.* ### PD Satellite Models — Estimation (Appendix V, Table 1) - Note: * indicates P value less than 5 percent. All dependent variables are confirmed to be stationary under Dickey-Fuller test. - Models and reported metrics: - PD household (Logit), PD corporate (Logit), Coverage ratio (Logit). - Coefficients and diagnostics (selected): - Unemployment rate, percent: PD household 0.18*, PD corporate 0.35*, Coverage ratio 1.23 - Short term rate, percent: Coverage ratio 1.34 - Term premium, percentage point: Coverage ratio 0.3* - Goodness-of-fit and sample size: - R square: 0.34, 0.75, 0.11 - Root mean square error: 0.9, 1.11, 4.9 - Number of observations: 78, 78, 35 ### Interest Rate Satellite Models — Liabilities (Appendix VI, Table 1) - Note: * indicates P value less than 5 percent. All dependent variables are confirmed to be stationary under Dickey-Fuller test. - Dependent series (percent): Overnight retail deposits; Overnight wholesale deposits; Overnight wholesale FX deposits; Indexed term deposits; Non-indexed term deposits. - Selected coefficient highlights: - Unemployment rate, percent: Overnight retail deposits 0.11* - Short term rate, percent: Overnight retail deposits 0.39*, Overnight wholesale deposits 0.24*, Overnight wholesale FX deposits 0.16* - Long term rate, precent: Overnight wholesale FX deposits 0.73*, Indexed term deposits 1.42* - Exchange rate, yoy, percent: Indexed term deposits 0.05* - GFC dummy: coefficients 0.43, 0.52, 0.74 (across some series) - Diagnostics and sample sizes (as displayed): - R square: 0.67, 0.66, 0.60, 0.79, 0.83 - Root mean square error: 0.94, 0.63, 0.52, 0.87, 1.66 - Number of observations: 107, 107, 87, 107, 95 ### Interest Rate Satellite Models — Assets (Appendix VI, Table 2) - Dependent series (percent): Indexed mortgages; Non-indexed mortgages; Corporate loans; Consumer loans. - Selected coefficient highlights: - Inflation, percent: Indexed mortgages -0.09* - Short term rate, percent: Indexed mortgages 0.55* - Long term rate, precent: Corporate loans 0.34* - Indexed term deposit, percent: Indexed mortgages 0.66*, Non-indexed mortgages 0.26* - Non-indexed term deposit, percent: Indexed mortgages 0.93*, Non-indexed mortgages 0.88*, Corporate loans 0.38 - GFC dummy: Consumer loans 0.83* - Diagnostics and sample sizes: - R square: 0.81, 0.99, 0.96, 0.89 - Root mean square error: 0.89, 0.58, 0.96, 1.7 - Number of observations: 78, 78, 78, 78 ### Financial Soundness Indicators — Regional Comparison (Appendix VII) - Indicators presented (latest available until end-2022), by country: Denmark, Finland, Iceland, Norway, Sweden. - Charts include T1 capital ratio (In percent), NPL Ratio (In percent), Return on Assets (In percent), Liquid Assets to Total Assets (In percent), Cost to Income (In percent), Liquid Assets to Short Term Liabilities (In percent). - (Charts and numeric axes shown in the source.) ### Domain: Top-down Stress Test by FSAP Team — Banking Sector (Solvency Risk) — Assumptions 1. Institutional perimeter - Institutions included: Top three commercial banks (under IFRS9). - Market share: The top three commercial banks account for about 95 percent of the deposit taking corporations (excl. central bank) assets. - Data source and baseline date: - Supervisory data provided by the Central bank of Iceland. Other data sources include public sources (EBA Transparency Exercises, Banks’ Annual Reports, Statistics Iceland), commercial databases (Fitch, Haver Analytics), IMF Global Assumptions (GAS) and IMF WEO. - Data as of October 2022. - Consolidated at national bank level. 2. Channels of risk propagation - Methodology: Balance sheet-based tool developed by MCM. Satellite models developed by the FSAP team. - Satellite models for macro-financial linkages: - Credit risk: Parameter (PD, LGD, EAD) projections generated by product; modeling relies on IFRS9 modeling and transition matrices; analysis uses as starting points the PDs and LGDs reported by banks. - Net Interest Income: Two approaches (structural and empirical); empirical approach uses regression models and pass-through estimates; structural model uses repricing ladders. - Net Fees and Commission income and other income/expenses: bank-panel model or by assumption. - Market risk: Duration approach for interest rate instruments and consideration of equity, FX and inflation risks. - Stress test horizon: 5 years (2023-2027). 3. Tail shocks / Scenario analysis - Baseline from the March 2023 WEO, complemented with VAR model to project scenario consistent additional variables. - Adverse scenario calibrated to a 2.2 standard deviation shock to real GDP growth relative to baseline over 2023-2024. - Macro-financial simulations realized based on an Iceland-specific VAR model and benchmarked against the “other advanced economies” group dynamics. - Macro-financial scenarios for foreign countries and relevant interest rates rely on the GFM simulations. 4. Risks and buffers - Risks/factors assessed: Credit risk (corporates, households, sovereign); Interest rate risk in the banking book; Market risk from fixed income securities (interest rate, spreads), FX inflation and equity risks. - Behavioral adjustments: - Balance sheet assumptions such that credit growth ensures that credit to GDP ratio remains constant. - New credit production endogenously consistent with credit growth assumption. - No write-off or cure allowed over projection horizon. - Hedging strategy is not considered. 5. Regulatory and market-based standards and parameters - Calibration of risk parameters: - PDs and LGDs obtained from supervisory files, or where not available estimated at the asset class level. - Dynamics based on model estimated PDs in line with the scenario considered (WEO baseline, adverse scenarios). - Projections of PDs and the starting positions of the transition matrices were used to guide shifts of transition matrices over the risk horizon to calculate loan loss provisions under the IFRS9 approach. - Regulatory/accounting and market-based standards: Regulatory capital ratios and IFSR9 accounting standards. 6. Reporting format for results - Output presentation: Aggregate results (regulatory capital and leverage ratios) and contributions to evolution of capital ratios. ### Banking Sector: Solvency Risk — Additional modelling details - Adverse scenario characteristics: - U-shaped path for real GDP growth, tightening of global financial conditions, global supply chain disruptions, rise of commodity prices, a de-anchoring of inflation expectations and a trade-off for monetary policy between unemployment and inflation. - VARX specification: - Endogenous vector Y_t includes real GDP, unemployment rate, CPI index, policy rate, nominal effective exchange rate (NEER), and a measure of real loans (CPI deflated sum of NFC and household loans). - Exogenous vector X_t includes US real GDP, oil price, and a GFC dummy (2008Q4). - Lag lengths: L_1 and L_2 chosen to be L_1=L_2=4. - All variables, other than unemployment and policy rates, are in logs. - Path of external variables taken from the global model maintained by MCM. ### Banking Sector: Liquidity Risk — Assumptions and Methodology 1. Institutional perimeter - Institutions included: Top three commercial banks. - Market share: The top three commercial banks account for about 95 percent of the deposit taking corporations (excluding central bank) assets. - Data and baseline date: - LCR, NSFR, and Cash flow table from supervisory data. - Data as of October 2022. - Consolidated at national bank level. 2. Channels of risk propagation - Methodology: - Cash-flow stress test analyzes net cash balance, accounting for available unencumbered assets, contractual cash inflows and outflows, and behavioral flows. - Repeated for all significant currencies for reporting banks. - Complemented with LCR and NSFR stress tests. - Stress test horizons: - Cash-flow analysis: horizon of stress events varies by scenario and can extend up to a period of 12 months. - LCR stress test horizon: one month. 3. Tail shocks and sensitivity analysis - Baseline and various adverse liquidity scenarios considered. - Sensitivity analysis includes further withdrawal of funding from pension and foreign funding. 4. Risks and buffers - Funding liquidity risk reflected in funding and asset roll-off rates. - Market liquidity risk reflected in asset haircuts, informed by market value declines from solvency stress tests and ECB valuation haircuts. - Behavioral adjustments: - Liquidity from central bank ELA is not considered. - Cash-flow analysis may consider behavioral assumptions about counterparties’ willingness to transact. 5. Regulatory and market-based standards - Calibration of stress funding run-off rates, asset roll-over rates, and asset haircuts based on empirical evidence and international experience. - LCR hurdle rate set at 100 percent at aggregate currency level and at 100 percent for significant foreign currencies. - NSFR per Basel III; limit of 100 percent. 6. Reporting format for results - Outputs include: (1) Changes in the system-wide liquidity position, (2) number of institutions with LCR/NSFR below regulatory limits, and (3) amount of liquidity shortfall. 7. Infrastructure - Infrastructure developed by IMF staff based on FINREP/COREP data input. ### Interconnectedness Analysis — Banks, Pension Funds, Investment Funds (Appendix IX) 1. Institutional perimeter and data - Interbank network: 3 commercial banks (out of 4) accounting for 95 percent of total banking sector assets, ranked by unconsolidated assets. - Inter-pension fund network: largest 15 pension funds ranked by total assets. - Inter-Investment fund network: largest 15 investment funds ranked by total assets. - Data granularity: institutional level bilateral exposure data among all entities; cross-border data for banking sector and pension funds at institutional level. - Starting position: three snapshots: 2011, 2017, and 2022. - Financial market data: sovereign CDS spreads and equity returns data from 2001 to 2022. 2. Methodology - Interbank and cross-border balance sheet exposure based on Espinosa-Vega and Juan Sole (2010). - Failure thresholds institution-specific. - Cross-border: market price-based spillover model by Diebold and Yilmaz (2014). 3. Risks and buffers - Risks considered: credit shock and funding shock causing capital impairment via interbank and intra-financial exposures. - Buffers: institutions’ own capital and liquidity buffers; Banks: minimum CET1 ratio considered; Pension funds: minimum solvency capital ratio. 4. Reporting format for results - Outputs include domestic and cross-border interconnectedness and contagion analysis, network charts, index of vulnerability and contagion, distribution of spillover indices, market data contagion analysis, and spillover indices at country level on sovereign CDS and equity return. ### Pension Funds: Stress Testing Matrix (Appendix X) — Top-down and Bottom-up 1. Institutional perimeter - 7 occupational pension funds (defined ambition): Almenni, Birta, Frjalsi, Gildi, LSR, LV, Stapi. - Market share: 77 percent of Pillar II assets, 76 percent of Pillar II contributions; excl. closed defined-benefit schemes. - Reference date: December 2022. 2. Channels of risk propagation and time horizons - Methodology: Investment assets affected by market value changes after price shocks; liabilities affected by changing assumptions on wage inflation, asset returns, regulatory discount rates, and biometric assumptions. - Time horizon: - Adverse scenario: 2023-2025. - Medium- to long-term projections for future pension values (up to 30 years). - Instantaneous shock for some analyses. 3. Tail shocks (Adverse scenario) and sensitivities - Adverse scenario shocks in 2023: - Interest rates: short-term rates +171 bps, long-term rates +212 bps in 2023. - Equity price: -79.8 percent for listed domestic shares, and -32.0 percent for foreign shares in 2023. - ISK depreciation: -30.6 percent in 2023. - Inflation: 8.6 percent in 2023. - Sensitivity analysis: - Default of largest bank / non-financial counterparty. - Reduction in the discount rate from 3.5 to 3.0 percent. - Decrease in mortality by 10 percent across all age cohorts. 4. Risk factors assessed - Market risks: interest rates, share prices, property prices, FX rates, credit spreads. - Credit risks: Default of largest bank (and non-financial) counterparty. - Regulatory/interest rate risk; Biometric risks: Mortality. 5. Regulatory/accounting standards - National GAAP (Top-down and Bottom-up). 6. Reporting formats for results - Outputs: Impact on value of assets, impact on future pension values, dispersion across companies, contribution of individual shocks, impact on value of assets and liabilities. ### Non-Financial Corporate Stress Testing Matrix (Appendix XI) 1. Institutional perimeter - About 8,500 non-financial companies. - Market share: About 25 percent of active firms in 2020. - Data source and reference date: - Orbis (Bureau van Dijk) database for company level data. - Statistics Iceland for aggregate sectoral key indicators. - Data as of December 2020. 2. Methodology and horizon - Dynamic Scenario-Based Stress Tests and Sensitivity Analysis (Tressel, T. and Ding, X., 2021). - Probability of default (PD) modelling. - Time horizon: Instantaneous shock and 3 years (2021-2023). 3. Scenario analysis and sensitivities - Baseline scenarios aligned with bank solvency stress test and October 2022 WEO. - Adverse scenario: lower GDP growth consistent with the severity of bank solvency stress test, tightening financial conditions, global supply chain disruptions, and rise of commodity prices. - Sensitivity analysis includes interest rate shock. 4. Risks and buffers - Risks/factors assessed: Bankruptcy, default on loans or bonds, ICR falling below specific thresholds. - Behavioral adjustments: None. 5. Reporting format for results - Outputs: Aggregate results with the impact on debt distress, contribution of individual shocks. *Italicized source attribution: IMF staff calculations and appendices as presented in the source document.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2023/english/1islea2023003.pdf_