## EXECUTIVE SUMMARY (cr17261) ## Source details **Canonical URL:** [EXECUTIVE SUMMARY (cr17261)](https://www.imf.org/-/media/files/publications/cr/2017/cr17261.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2017/cr17261.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2017/cr17261.pdf.json) --- ### Financial system overview - Financial soundness indicators: "relatively robust." - Investment fund industry: - Assets under management: €3.6 trillion or 67 times GDP. - Industry composition: Bond funds 35 percent (€1,170 billion as of July 2016), Equity funds 33 percent, Mixed funds 26 percent, Other funds ~6 percent. - UCITS regime: majority of assets. - Banking sector: - Sample of 16 banks in TD solvency tests represents 73 percent of banking sector assets (excluding branches), 94 percent of residential mortgage loans, and 68 percent of exposures to foreign banks. - CET1 ratio (pre-stress, fully-loaded Basel III): 18.6 percent (16 banks); domestically-oriented 20.1 percent; internationally-oriented 17.8 percent. - Tier 1 ratio (pre-stress): 18.7 percent (16 banks). - Total Capital ratio (pre-stress): 19.4 percent (16 banks). - Leverage ratio (Tier 1 / total assets, pre-stress): 5.9 percent (16 banks). - Return on assets: 0.8 percent at end-2015. - Nonperforming loan ratio: 1.2 percent as of June 2016 (16-bank sample). - Loan-to-deposit ratio: 65 percent; liquid assets-to-short-term liabilities ratio: 68 percent; standard LCR metric at 80 percent in 2017 (all 16 banks meet by a comfortable margin). - Insurance industry: - Life insurers in sample: 10 life insurers covering 71 percent of technical provisions; share of unit-linked products in sample insurers: 77 percent. - Aggregate Eligible Own Funds decline by 27 percent in third year of adverse scenario relative to 2016 baseline, but all tested life insurers and the reinsurer remain above Solvency II minimum requirement. - Household sector: - HFCS microdata used (2014); pre-shock implied PD: 2.5 percent. ### Key structural vulnerabilities - Size and interconnectedness: - Investment fund AuM: €3.6 trillion or 67 times GDP. - Herfindhal index of the banking system: 5 percent (low concentration). - Main vulnerabilities: - Liquidity aggregation by foreign bank subsidiaries from investment fund and wealth management operations and upstreaming to parents abroad, exposing parents to maturity and currency transformation. - Investment fund liquidity buffers may struggle to accommodate large redemption shocks, potentially causing asset fire sales and drawdown of local bank deposits (operational deposits broadly stable in past stress periods). - Elevated real estate valuations and easing lending standards: main home-grown vulnerability; domestically-oriented banks’ exposures to domestic real estate make up 20 percent of domestically-oriented banks’ total assets. - Concentration exposures in funds: - Around half of sampled EM bond funds have exposures to single EM sovereign issuers >5 percent of TNA; some up to 30 percent of TNA. - UCITS HY funds account for 5 percent of total net assets (TNA) of Luxembourg’s investment fund industry. ### Stress test design, scope, and scenarios - Horizon and data cutoff: - Three-year horizon: 2017–2019; data through June 2016 (except fund data March 2016). - Tests conducted: - Banking sector: Top-down solvency stress test (TD) by FSAP team with macro scenarios and sensitivity analyses; TD liquidity stress tests (LCR, NSFR, cash-flow); contagion/network models. - Households: micro-data-based household stress test using HFCS. - Insurance: bottom-up solvency stress tests coordinated by CAA. - Investment funds: IMF staff solvency test for CNAV MMFs and liquidity stress tests for UCITS bond funds (sample of 191 funds totaling €656 billion; total TNA universe €1,336 billion; sample coverage 49.1 percent of TNA universe). - Investment fund sample coverage (selected): - MMF sample covered 75 percent of the sector. - Liquidity stress test coverage: 75 percent of High Yield (HY) and Emerging Market (EM) bond funds, 54 percent of mixed funds (fixed income focus), 32 percent of other bond funds. - Redemption shock calibration: - Historical approach: 1% highest monthly net flows observed (2007–2016) specific to each fund. - Forward-looking approach: econometric model linking strategy-level net flows to macrofinancial variables under the adverse macro scenario. - Adverse macro scenario calibration highlights: - Luxembourg GDP: -3.7, -2.6 and +1.8 percent in 2017, 2018 and 2019 respectively; average real GDP growth of -1.5 percent per annum over 2016–2018; cumulative shock over 2017–2019 of 14.7 percentage points. - House prices: cumulative decline of 30 percent over 3 years. - Investment funds' total assets: cumulative drop of 15 percent over first year. - Unemployment rate peak: 11 percent in second year. - Corporate cash bond spread shock (year 1 peak): 119 bps based on historic episode. ### Banking sector stress test results (solvency) - Aggregate outcome under severe adverse scenario: - Aggregate CET1 ratio declines from 18.6 percent (June 2016) to 15.1 percent (three-year horizon), a drop of 3.5 percentage points. - Contribution breakdown to CET1 decline: - Credit losses: -1.6 percentage points of RWAs. - Market losses in bond portfolio: -1.4 percentage points of RWAs. - Change in risk-weighted assets: -2.5 percentage points. - Offsetting widening interest margin and other effects: +2 percentage points. - Expected credit losses over three years in adverse scenario: €2.7 billion (0.6 percent of total banking system assets). - Baseline scenario new provisions: €0.6 billion (0.1 percent of total banking system assets). - Capital shortfalls: - Banks with capital falling below regulatory Capital Adequacy ratio of 8 percent: banks making up less than 10 percent of banking sector assets; recapitalization needs = 0.8 percent of GDP. - Based on CET1 hurdle rate of 4.5 percent, capital shortfall reduces to 0.1 percent of GDP. - Leverage ratio effects: - Tier 1 / total assets (leverage ratio) declines from 5.9 to 5.0 percent. - Six banks would fall below 3 percent hurdle in 2019, equivalent to 1.6 percent of GDP. - Valuation losses on AFS and HFT bond portfolios in adverse scenario: €2.5 billion (contributes 1.4 percentage points to CAR decline). - FX position: net open FX position €6 million as of June 2016 (0.02 percent of Tier 1 capital); FX losses negligible (e.g., €0.4 million in adverse scenario). - Sensitivity tests (static) example shocks: - Lending rate shock: severely adverse 86 bps; sensitivity test 500 bps. - Domestic sovereign interest rate shock: severely adverse 177 bps; sensitivity test 500 bps. - FX shock: severely adverse -7.4 percent; sensitivity test +/- 30 percent. - Equity price shock: severely adverse -19 percent; sensitivity test -50 percent. - Concentration stress results: - Simultaneous default of five net largest exposures → three banks undercapitalized at Tier 1 6 percent; capital shortfall = 1.6 percent of GDP. - Simultaneous default of ten net largest exposures → six banks undercapitalized; capital shortfall = 8.9 percent of GDP. - Reintegration of exempted corporate exposures and zero recovery assumption ("double default") increases shortfalls up to 24.4 percent of GDP in extreme scenarios. ### Household stress test results - Scenario: cumulative drop in house prices of 30 percent over three years; unemployment +5 percentage points; annual wages and unemployment benefits drop by 5 percent. - Implied households' PDs: - Current implied PD: 2.5 percent. - Peak implied PD under the scenario: 6.4percent. - Liquidity buffers: - More than 93 percent of indebted households have sufficient liquid assets to cover at least 12 months of income shortages. - Share of outstanding mortgage balance for 1st decile (highest PD): 4.1 percent of total outstanding mortgage balance. - Interpretation: households’ solvency significantly impacted but proportion of households in default would remain manageable given high household financial wealth. ### Bank liquidity stress tests and contagion findings - LCR-based results (aggregate, September 2016): - Aggregate LCR = 109.4 percent under standard scenario. - Four out of 16 banks do not pass 100 percent hurdle (binding in 2018); shortfall at 100 percent hurdle = €3.5 billion (0.8 percent of banks' assets). - Luxembourg retail stress scenario (20–30 percent retail run-offs): - Aggregate LCR = 107.6 percent. - Liquidity shortfall = €1.9 billion (0.4 percent of banks' assets). - Luxembourg wholesale stress scenario (50–100 percent unsecured wholesale run-offs): - Aggregate LCR = 96.3 percent. - Seven banks below 100 percent; two below 80 percent. - Total liquidity shortfall = €7.2 billion (1.6 percent of banks' assets). - Foreign-currency LCR tests: liquidity shortfalls range €0.3 billion to €12.7 billion (0.1 percent to 3.5 percent of banks' assets); USD positions particularly weak. - NSFR (June 2016): - Available stable funding = €165 billion; Required stable funding = €160 billion; Aggregate NSFR = 101.1 percent. - Two banks fall short (authorities expect compliance by January 2018). - Cash-flow (outflow) analysis: - Analysis covered eight banks (45 percent of banking assets); all but one bank could meet 1–7 day gaps by selling liquid assets equaling 9.2 percent of total assets. - Reliance on securities to fund gaps is a vulnerability if multiple banks sell simultaneously. - Contagion/network findings: - Domestic interbank exposures small relative to capital; no single domestic failure would trigger another bank failure in sample. - Limited direct contagion via bank–insurer cross-exposures. - Significant cross-border intragroup exposures: weighted average exposure across banks almost 6 times subsidiary capital (exceeds typical waiver norms); combined sum of large exposures: €362 billion; combined funding from 10 largest counterparties: €220 billion (162 billion to credit institutions; ~140 billion intra-group bank-to-bank funding). - Market-based network analysis: Luxembourg banks not strongly connected domestically; investment funds closely connected to each other and to global banks; two banks show close links to bond and mixed funds. ### Investment fund liquidity and MMF findings - Fund sample: 191 funds totaling €656 billion; TNA universe €1,336 billion; sample coverage 49.1 percent of TNA. - Redemption shocks (percent of TNA): - Historical approach averages: EM 18%, HY 19%, Mixed 9%, Other bond funds 18%, Short Term CNAV 19%, Short Term VNAV 23%, Other MMFs 18%. - Forward-looking (adverse macro) approach: EM 9%, HY 11%, Mixed net inflows 11%, Other bond funds 6% (MMFs excluded from forward-looking models). - Liquid asset measures: - Short-term assets: cash + debt securities residual maturity < one year. - HQLA approach: ESMA (2015)-based liquidity weights by rating, instrument, issuer; non-investment grade corporate bonds weight = 0 percent. - Liquidity stress-test outcomes: - Under HQLA measure: most fixed income funds resilient to severe redemption shocks; HY funds more vulnerable. - Under short-term asset measure: MMFs and most mixed funds resilient; other funds may face liquidity shortfalls. - More than half of HY funds reported insufficient liquid assets under both scenarios and both liquidity measures. - Deposit outflow estimates from funds (data caveat: assumes 100 percent of fund deposits with their depositary bank): - Estimated deposit outflows ~25 percent of total cash deposits of funds in sample ≈ €5.2 billion. - Incorporating projected mixed fund inflows: close to 20 percent ≈ €4.9 billion. - Alternative waterfall/prorata approaches produce deposit outflow averages (in EUR billion and percent of fund deposits): - Total deposits including mixed funds: 24.7; deposit outflows (average) 4.9 (20% of fund deposits). - Total deposits without mixed funds: 21.6; deposit outflows (average) 5.3 (24% of fund deposits). - MMF solvency ("break the buck") tests for CNAV MMFs: - Shadow NAV deviation of 0.2 percent from par would require combined 170 basis point increase in risk-free rates and 70 basis point rise in credit spreads (asset-weighted average of five largest CNAV MMFs). - CNAV holdings: deposits and reverse repo ≈ one quarter of portfolio; short-term debt instruments >50% of holdings. - Historical largest shadow NAV deviation for 10 largest CNAV MMFs: below 20 basis points except one slightly above 20 basis points; no suspensions/gates reported. - Fund concentration findings: - Around half of EM bond funds have single-issuer exposures >5 percent TNA; some funds up to 30 percent of TNA to single EM sovereigns. - Bond and mixed funds: 34 and 46 percent of sample TNA have large exposures to individual sovereigns respectively. ### Aggregate implications and supervisory priorities - System resilience: high starting capital and liquidity allow most banks and life insurers to absorb large market shocks while retaining buffers; money market fund solvency and liquidity risks appear muted at aggregate level. - Main supervisory priorities: - Monitor investment fund liquidity risks and bank–fund interlinkages. - Address data gaps on funding sources, cross-border bilateral liabilities, and short-term maturity information. - Strengthen oversight of intragroup exposures and FX liquidity risk at group level. ### Key near-term (NT) and medium-term (MT) recommendations (excerpt) - NT: Ensure banks’ Internal Ratings-Based models take into account losses in severely stressed environments. — ECB/CSSF. - MT: Consider implementing a regulatory LCR requirement in foreign currencies at the group level and step up monitoring of related FX liquidity risk. — European Commission, ECB. - MT: Work with the EBA to close liquidity reporting gaps and expand harmonized EU bank reporting. — CSSF/EBA. - NT: Extend maturity of wholesale funding, including deposits by parents. — ECB/BCL/CSSF. - NT: Intensify bank liquidity supervision of foreign currency mismatches and perform liquidity stress tests based on cash flows at various maturities. — ECB/CSSF. - NT: Conduct periodic reviews of bank eligibility for large intragroup exemption waivers. — ECB/CSSF. - NT: Improve quality of metadata reporting on large exposures. — ECB. - NT: Close data gaps on funding sources by expanding scope beyond the 10 largest counterparties. — ECB. - NT: Provide industry guidance on liquidity stress tests for investment funds exposed to less liquid asset classes and develop internal liquidity stress testing capacity. — CSSF. - NT: Commission a study on the effectiveness of investment fund liquidity management tools to support industry guidance. — CSSF. - NT: Continue assessing and monitoring bank–fund interlinkages, especially for depositary banks. — CSSF, BCL. - NT: Intensify monitoring of concentration risk, particularly single sovereign issuer concentrations. — CSSF. *Source: EXECUTIVE SUMMARY (cr17261).* ### EXECUTIVE SUMMARY ___________________________________________________________________________ 7 ### EXECUTIVE SUMMARY ___________________________________________________________________________ 7 ### INTRODUCTION - The document contains an introductory section located on page 13 that frames the scope and aims of the assessment. ### SOLVENCY STRESS TESTS - Section begins on page 18 and is organized into four subsections: - A. Solvency Stress Test of the Banking Sector (page 18) - Includes calibration of solvency stress tests and references to FSAP stress test hurdle ratios (Table 2). - Contains bank-level risk indicators and solvency test results (Figures 3, 4, 5, 6, 7, 8, 9, 10, 11, 12; Tables 2, 3, 4). - B. Household Stress Testing Methodology and Results (page 39) - Technical details on household stress tests are provided in Appendix VI. - C. Solvency Analysis and Stress Test of the Investment Fund Sector (page 40) - Investment funds coverage and concentration analyses are presented (Figure 13; Table 5). - CNAV money market fund portfolios and MMF solvency technical details are included (Figures 14; Appendix VII). - D. Solvency Stress Test of the Insurance Sector (page 46) - Life insurers’ bottom-up stress test results are summarized (Figure 15). ### LIQUIDITY STRESS TESTS - Section begins on page 47 and includes two subsections: - A. Liquidity Stress Tests for the Banking Sector (page 47) - LCR-based stress test assumptions and results are presented (Table 6; Figure 16; Table 7). - B. Liquidity Risks in the Investment Fund Industry (page 54) - Outflow analysis and liquidity coverage for UCITS and other funds (Figures 17, 18, 19; Tables 8, 9, 10). - Box 1 and Box 2 provide technical methods for redemption shock estimation and liquidity asset measures (pages 55–56). - Appendix VIII contains technical details on the liquidity stress test for UCITS funds. ### INTERCONNECTEDNESS ANALYSIS AND CONTAGION RISKS - Section begins on page 61 and includes: - A. Domestic Interconnectedness Analysis (page 61) - Interbank network analysis and network-model implementation (Figures 20; Boxes 3 and 4). - B. Cross-border Contagion Risks (page 65) - Cross-border banking network analyses and intragroup exposures (Figures 22, 23, 24, 25, 26). - Market-based network analyses covering 2011–2016 (Figure 27; Box 5). ### CONCLUSION - Concluding remarks are located on page 76. ### SUPPORTING MATERIALS, FIGURES, TABLES, BOXES, AND APPENDICES - Boxes: 1–5 covering redemption shocks, liquidity measures for funds, and technical summaries of network and market models (pages 55, 56, 69, 70, 74). - Figures: 1–27 covering asset management industry overview, stress test summaries, bank FSI, scenario calibrations, fund liquidity, network analyses, and more (listed with page references in the contents). - Tables: 1–10 including "Recommendations on Risk Analysis" (Table 1), FSAP hurdle ratios (Table 2), investment fund coverage (Table 5), liquidity stress test assumptions and results (Tables 6–10), and other supporting tables. - Appendices I–V and VI–VIII contain methodological details, stress test matrices, technical model descriptions, and sector-specific technical appendices (pages 78, 80, 85, 91, 94, 100, 104, 105). ### GLOSSARY AND ABBREVIATIONS - A comprehensive glossary of acronyms and terms is provided (pages following the main contents), including entries such as: - AIF, AFS, AuM, BCL, CAA, CCP, CDS, CET1, CNAV, COREP, CRD, CRR, CSSF, EBA, ECB, EDF, EL, ELA, EU, EUR, FDI, FINREP, FSGM, FSSA, FX, GBP, HFCS, HFT, HQLA, HY, IBB, ICAAP, IMF, LMT, LSI, NAV, NFC, NSFR, OLS, OTC, O-SII, PIT, PoD, RAM, RCR, RWA, SA, SI, STA, STE, STeM, TN, TNA, TTC, U.K., URB, U.S., USD, UCITS, VAR, VIX. - Page numbering and placement of the glossary entries are part of the document front/back matter. *LUXEMBOURG INTERNATIONAL MONETARY FUND 7* ### EXECUTIVE SUMMARY ### EXECUTIVE SUMMARY ### Financial system overview - Financial soundness indicators have remained relatively robust. - Investment fund industry: strong growth in assets under management amid rising asset prices and inflows; exposure to liquid assets has remained steady. - Banking sector: high levels of profitability, capital, liquidity, and asset quality; a relatively large share of liquidity and revenues derive from private banking and fund management activities. - Insurance industry: relatively less exposed to guaranteed products than regional peers; adjusted well to the new regulatory regime, maintaining high profitability and capitalization levels. ### Key structural vulnerabilities - Size and interconnectedness of the financial system create potential vulnerabilities. - Foreign bank subsidiaries may aggregate liquidity from investment fund and wealth management operations and upstream it to parents abroad, exposing parents (where oversight by Luxembourg authorities may be limited) to maturity and currency transformation. - Risk that investment fund liquidity buffers may struggle to accommodate large redemption shocks, potentially causing asset fire sales and a drawdown of local bank deposits (operational deposits have been broadly stable in past stress periods). - Elevated real estate valuations and easing lending standards pose the main home-grown vulnerability, especially for domestically-oriented banks. ### Stress test design and scope - Covered banks, insurance companies, investment funds, and households. - Banking sector stress tests: top-down (TD) exercise based on macroeconomic scenarios and sensitivity analyses over a three-year horizon (2017-2019), using data through June 2016; conducted by the FSAP team in cooperation with the ECB. - Credit risk parameter projections benchmarked to international experience from small open advanced economies with large financial systems that experienced severe banking crises. - TD liquidity stress tests: cash flow-based analysis and supervisory information on aggregate basis and by currencies. - Contagion tests: domestic interbank gross exposures, interbank cross-liabilities, bank-insurer cross-exposures, cross-border interlinkages using supervisory data; market data analysis for bank–fund and global bank linkages. - Household stress tests: micro-data-based shocks to income, home prices, and unemployment to benchmark banks’ mortgage credit loss projections. - Insurance solvency stress tests: bottom-up approach under guidance of the Commissariat aux Assurances (CAA), shocking security holdings and fee income from unit linked products. - Investment fund stress tests (IMF staff): solvency stress test for a sample of money market funds (MMFs) and liquidity stress tests for UCITS bond funds (Emerging Market, High Yield, mixed funds, and other bond funds). - Fund sample: 191 funds, totaling €656 billion in total net assets. - MMF sample covered 75 percent of the sector. - Liquidity stress test coverage: 75 percent of High Yield and Emerging Market bond funds, 54 percent of mixed funds investing mainly in fixed income instruments, and 32 percent of other bond funds in Luxembourg. - Redemption shock calibration: historical approach (worst month investor outflows specific to each fund) and forward-looking approach (based on adverse macroeconomic scenario used in bank stress tests). ### External and domestic risk drivers in scenarios - Key external risks: renewed euro area recession, surge in financial market volatility, and financial system outflows affecting external demand and investment fund flows; higher volatility could raise interest rates, credit spreads, and funding costs. - Key domestic risks amplifying adverse scenario severity: domestic confidence shock reducing consumption and investment, house price drop triggering adverse wealth effects, sharp drop in government revenue leading to sovereign rating downgrade and increased sovereign risk premia. ### Banking sector stress test results - Aggregate resilience: banking system remains very well-capitalized despite a meaningful decline in the system-wide capital ratio under the severe adverse scenario. - CET1 ratio: declines from 18.6 percent to 15.1 percent over the three-year period. - Domestically-oriented banks are more severely impacted than internationally-oriented banks. - Main vulnerabilities: credit risk in the corporate portfolio and market losses in bond portfolios, especially foreign sovereign bond and non-sovereign portfolios (without possible macro hedges). - Capital shortfalls: - A number of banks (making up less than 10 percent of banking sector assets) would see capital fall below the minimum regulatory Capital Adequacy ratio of 8 percent; recapitalization needs would amount to 0.8 percent of GDP. - Based on the CET1 hurdle rate of 4.5 percent, the capital shortfall would be reduced to 0.1 percent of GDP. ### Household stress test results - Households’ solvency would be significantly impacted by declines in income and home prices and a rise in unemployment, but the proportion of defaulted households would remain manageable, reflecting relatively large financial wealth of Luxembourgish households. ### Bank liquidity and contagion findings - Bank liquidity: broadly resilient but would be weakened if unsecured wholesale funding dries up or if funding stress emerges in foreign currencies. - Data limitations: parts of liquidity stress tests were restricted to only eight banks, creating substantial gaps in liquidity analysis. - Under a more severe scenario than Basel III LCR metrics: some banks would be exposed to liquidity risks in an extreme dry-up of unsecured wholesale funding including operational deposits by investment funds. - Some banks show exposures to group entities because foreign currency funding is channeled to foreign parents. - Reliance on securities to fund large short-term gaps could be a vulnerability when selling large asset quantities risks reputational losses or large haircuts. - Contagion analysis: limited direct contagion risk through domestic bilateral exposures and bank–insurer cross-exposures; significant cross-border bank linkages due to Luxembourg’s role as hub for foreign parent bank subsidiaries. - Market data analysis: Luxembourg banks are not strongly connected domestically; investment funds are closely connected to each other and to global banks. ### Investment fund liquidity and MMF results - Liquidity measures: - Most restrictive: cash and short-term securities debt (maturity less than one year). - HQLA approach: attributes liquidity weights by rating, instrument type, and issuer (following ESMA (2015)). - Liquidity stress test outcomes: - Most fixed income funds would be resilient to severe redemption shocks when liquidity buffers measured by HQLA; HY funds more vulnerable than others. - MMFs and most bond funds have sufficient liquid assets to meet severe redemption shocks under the HQLA approach. - Forward-looking (macro scenario) approach: mixed funds see inflows; EM and other bond funds have sufficient liquid assets to avoid selling less liquid assets or using liquidity management tools. - Under the restrictive cash and short-term debt measure: MMFs and most mixed funds would be resilient; other funds would face liquidity shortfalls. - More than half of HY funds reported insufficient liquid assets under both scenarios and both liquidity measures, though HY fund manager self-assessments via CSSF UCITS risk reporting indicate a more benign view. - Market shares: UCITS HY funds account for only 5 percent of total net assets (TNA) of Luxembourg’s investment fund industry. - Redemption shock transmission: under the adverse scenario, funds exposed to less liquid asset classes would face liquidity shortfalls that could cause deposit outflows from banks; estimated fund deposit withdrawals would range between 20 and 25 percent of fund deposits in the sample. - Note on estimation: due to data gaps, deposit outflows are estimated assuming 100 percent of funds’ deposits are with their depositary bank, while in practice deposits are likely spread across several banks inside and outside Luxembourg. ### MMF solvency ("break the buck") tests - CNAV MMFs: very large increases in interest rates and credit spreads would be needed for 'shadow' NAVs to depart significantly from par value. - Illustration: a 0.2 percent deviation between shadow NAV and par value would require a combined 170 basis point increase in risk free rates and a 70 basis point rise in credit spreads. ### Aggregate implications and supervisory priorities - Case for further monitoring of investment fund liquidity risks and bank–fund interlinkages. Recommended actions include: - Regular liquidity stress tests for funds exposed to less liquid asset classes (HY and EM bond funds). - Provide industry guidance on liquidity stress test modalities (coverage, frequency, scenarios) to ensure consistency and comparability. - Study effectiveness of investment fund liquidity management tools to inform subsequent industry guidance. - Heightened supervisory scrutiny of concentration risks, particularly large exposures to single sovereign issuers. - Continued monitoring of bank–fund interlinkages by CSSF and BCL through the Comité du Risque Systemique (CSR) workstream. ### Key near-term and medium-term supervisory recommendations (excerpted) - Ensure banks’ Internal Ratings-Based models take into account losses in severely stressed environments. (NT) — ECB/CSSF. - Consider implementing a regulatory LCR requirement in foreign currencies at the group level and step up monitoring of related FX liquidity risk. (MT) — European Commission, ECB. - Work with the EBA to close liquidity reporting gaps and expand harmonized EU bank reporting. (MT) — CSSF/EBA. - Extend maturity of wholesale funding, including deposits by parents. (NT) — ECB/BCL/CSSF. - Intensify bank liquidity supervision of foreign currency mismatches and perform liquidity stress tests based on cash flows at various maturities. (NT) — ECB/CSSF. - Conduct periodic reviews of bank eligibility for large intragroup exemption waivers. (NT) — ECB/CSSF. - Improve quality of metadata reporting on large exposures. (NT) — ECB. - Close data gaps on funding sources by expanding scope beyond the 10 largest counterparties. (NT) — ECB. - Provide industry guidance on liquidity stress tests for investment funds exposed to less liquid asset classes and develop internal liquidity stress testing capacity. (NT) — CSSF. - Commission a study on the effectiveness of investment fund liquidity management tools to support industry guidance. (NT) — CSSF. - Continue assessing and monitoring bank–fund interlinkages, especially for depositary banks. (NT) — CSSF, BCL. - Intensify monitoring of concentration risk, particularly single sovereign issuer concentrations. (NT) — CSSF. *Source: EXECUTIVE SUMMARY (cr17261).* ### 2. Nevertheless, the defining structural characteristics of the financial system—size and ### 2. Nevertheless, the defining structural characteristics of the financial system—size and interconnectedness—as well as elevated real estate valuations, give rise to potential vulnerabilities. ### Key structural features and vulnerabilities - Herfindhal index of the banking system: 5 percent, indicating a low level of concentration. - Luxembourg is home to the world's second largest investment fund industry by assets under management: €3.6 trillion or 67 times GDP. - Main structural vulnerabilities identified: - Liquidity aggregation by Luxembourg foreign bank subsidiaries from investment fund and wealth management operations and ‘upstreaming’ to parents abroad may expose parents to maturity and currency transformation risks, with limited Luxembourg oversight when parents are not regulated under the Single Supervisory Mechanism. - Investment fund liquidity buffers could struggle to accommodate large redemption shocks, potentially triggering asset fire sales and a drawdown of local bank deposits (though operational deposits have been broadly stable in past stress periods). - Real estate market: strong run up in prices and easing in lending standards; domestically-oriented banks are most exposed and this poses the main home-grown vulnerability. ### Investment fund industry: size, composition, resilience, and vulnerabilities - Size and composition: - Total assets under management: €3.6 trillion or 67 times GDP. - Asset class breakdown (share and specific values where provided): - Bond funds: 35 percent (€1,170 billion as of July 2016 according to ECB data). - Equity funds: 33 percent. - Mixed funds: 26 percent. - Other funds (real estate, hedge and other): around 6 percent. - Distribution: Luxembourg investment funds are distributed to over 70 countries, with particular appeal in Europe and Asia. - Legal/regulatory framework: majority of assets fall under the UCITS regime. - Historical resilience: - Aggregate redemptions remained limited during the Global Financial Crisis and the European sovereign crisis due to diversification; a few funds suspended redemptions but the industry withstood major disruptions. - Main vulnerabilities of the investment fund industry: - Concentrated unhedged exposures can leave funds disproportionately affected by distress of particular issuers and difficult to liquidate in stress situations, especially if funds own a large share of individual bonds. - Daily liquidity offered by most Luxembourg funds can create liquidity mismatches if funds invest in less liquid assets (e.g., emerging market debt or HY bonds). If liquidity management tools are ineffective, a large redemption shock could force rapid liquidation, causing fire sales and price spirals. - Bank linkages: on the asset side, funds hold bank debt securities, bank deposits, and have derivatives exposures with banks as counterparties; on the liability side, funds may have overdraft facilities and credit lines with banks that may not be available during stress. - Potential systemic transmission: - A large redemption shock that exhausts fund liquidity buffers and forces portfolio adjustments could lead to bank deposit withdrawals, predominantly affecting international banks providing depositary services. ### FSAP stress testing objectives, scope, and methodology - Objective: - Assess capacity of the financial system to withstand severe but plausible macroeconomic shocks; explore weaknesses and transmission channels; identify policy priorities (reducing exposures, building capital and liquidity buffers); identify informational and methodological gaps and assess preparedness. - Distinctions and caveats: - FSAP stress tests differ from microprudential supervisory stress tests; FSAPs have a macrofinancial perspective and are not intended to prompt supervisory actions at individual-bank level. - Results are subject to data and methodological limitations (no reflection of possible economic hedges, use of benchmarks from other countries due to lack of relevant historical data). - Adverse scenarios are combinations of external and domestic shocks that ex ante have a very low probability of occurrence; they should not be interpreted as macroeconomic forecasts. - Consistency: - Consistency is sought across FSAPs in methodology, though country-specific environments and methodological differences limit full comparability. ### Stress test design, coverage, and scenarios - Overall test types applied: - Solvency, liquidity, and contagion tests for the banking sector (Top-Down (TD) by FSAP team with macro scenarios and sensitivity tests). - Household stress test using HFCS microdata. - Insurance sector bottom-up solvency stress tests coordinated by the CAA using financial market shocks from the macro-financial model. - Investment fund tests: solvency test for CNAV MMFs and liquidity stress test for five types of fixed income funds. - Macroeconomic scenario horizon and data cutoff: - Three-year horizon: 2017–2019. - Based on data available through June 2016. - Banking sector coverage: - TD stress test based on sixteen banks: fifteen significant institutions (SIs) directly supervised by the ECB, and one less significant institution (LSI). - These 16 banks represent: - 73 percent of the banking sector's total assets (excluding branches of foreign banks). - 94 percent of residential mortgage loans. - 68 percent of exposures to foreign banks. - Liquidity stress tests: - LCR analysis covered 16 banks. - NSFR and cash-flow based stress tests covered only eight banks (45 percent of banking sector assets) due to data availability. - Investment fund coverage: - TD stress test covered 191 funds (all UCITS except one). - Representativeness: - 75 percent of the assets of High Yield (HY) funds, Emerging Market (EM) bond funds, MMFs. - 54 percent of the assets of mixed funds investing primarily in fixed income instruments. - 32 percent of other bond funds. - Insurance sector coverage: - Bottom-up solvency tests included 10 life insurers comprising 71 percent of technical provisions. - The one dominant reinsurer accounting for two thirds of sector assets was included. - Specific stress test features: - Bond funds and MMFs emphasized for liquidity stress tests due to size and susceptibility to liquidity mismatches. - For CNAV MMFs: assessed likelihood of a 'break the buck' scenario by subjecting portfolios to sharp increases in risk-free rates and credit spreads. - Investment fund liquidity stress tests calibrated redemption shocks on: - i) 1% highest historical net flows and - ii) the adverse macro scenario. - Liquidity assets measured by cash, short-term securities and HQLA. ### Banking solvency stress tests: regulatory framework and hurdle ratios - Regulatory frameworks referenced: - Basel III framework, Fourth Capital Requirements Directive (CRD IV), Regulation on Prudential Requirements, national law and CSSF regulation. - Hurdle rates set according to Basel III fully-loaded definitions; leverage ratio requirement considered (binding from 2018 onwards). - Capital and buffer treatment: - Stress tests based on minimum capital ratios under Pillar I; individual Pillar II requirements not taken into account except for dividend distribution policy assumptions. - Banks allowed to deplete a capital conservation buffer of 2.5 percent in the adverse scenario. - Most deferred tax assets and all goodwill were already fully deducted from Common Equity Tier 1 at the cutoff date of the FSAP (June 2016). - Unrealized capital gains were added to the initial capital base in anticipation of Basel III full implementation. - Hurdle ratios and actual ratios before stress (Table 2) - Total Capital ratio (total capital to RWAs, in percent) - Stress test hurdle rate (excluding capital conservation buffer): 8 - Actual ratio before stress (excluding unrealized capital gains): - 16 banks: 19.4 - Domestically-oriented: 20.8 - Internationally-oriented: 18.6 - Tier I Capital ratio (Tier 1 capital to RWAs, in percent) - Stress test hurdle rate: 6 - Actual ratio before stress: - 16 banks: 18.7 - Domestically-oriented: 20.4 - Internationally-oriented: 17.8 - Common Equity Tier I Capital ratio (CET1 capital to RWAs, in percent) - Stress test hurdle rate: 4.5 - Actual ratio before stress: - 16 banks: 18.6 - Domestically-oriented: 20.1 - Internationally-oriented: 17.8 - Leverage ratio (Tier 1 capital to total assets, in percent) - Stress test hurdle rate: 3 - Actual ratio before stress: - 16 banks: 5.9 - Domestically-oriented: 7.6 - Internationally-oriented: 5.1 ### Stress test modalities and transmission channels examined - Banking sector stress tests: - Top-down macro tests: external and domestic shocks; forecasts of credit losses and other profit/loss sources based on satellite models. - Sensitivity tests: domestic shocks and market risk sensitivity analyses. - Liquidity tests: LCR-type, NSFR-type, cash-flow based maturity ladder analysis; reverse liquidity sensitivity tests; currency breakdowns. - Contagion tests: pure interbank, banks-insurers, and cross-border contagion models; market-data-based network model inspired by Goodhart-Segoviano (2009) and Diebold-Yilmaz (2012) connectivity methodology. - Integration across sectors: - Common macro scenario used to integrate solvency tests across banks, insurers, and investment funds. - Consistent assumptions across stress test streams where applicable. *Source: IMF staff.* ### 26. The Luxembourgish financial system is exposed to several external risks. The risks that ### cr17261 - 26. The Luxembourgish financial system is exposed to several external risks. The risks that ### External risks most likely to materialize - A sharp rise in global financial market risk premia and flight to safety: domestic effects through lower investment fund and bank profitability; impact could be somewhat limited if Luxembourg acts as a recipient of safe haven inflows. - Financial system outflows resulting from a larger than expected hit to bank and investment fund exposures, in an environment of prolonged negative interest rates and rising regulatory demands. - Structurally weak (worse than currently projected) growth in key advanced and emerging economies, particularly the euro area, transmitted through extensive trade and capital flow channels. ### Domestic risks: real estate and household indebtedness - Domestically-oriented banks’ exposures to the domestic real estate market make up 20 percent of domestically-oriented banks’ total assets. - High loan-to-value ratio loans have accounted for a significant share of new mortgages; where such loans are granted additional collateral is generally required. - ESRB (November 2016) highlighted risks related to rising indebtedness and ability of households to repay mortgage debt and residential real estate price dynamics; noted loan-to-value and debt-service-to-income ratios in Luxembourg are moderate in the stock of existing mortgages and may act as a mitigating factor. ### Banking sector: current strengths (Figure 3) - Profitability: return on assets was 0.8 percent at end-2015, having risen back to pre-crisis levels; net interest margin compression has been less problematic due to reliance on fee and commission income from private banking and fund management. - Capitalization: Tier 1 capital represents 95 percent of total regulatory capital; most banks already meeting fully-loaded Basel III capital requirements; no bank found to have a capital shortfall by the 2014 ECB Comprehensive Assessment. - Asset quality: nonperforming loan ratio close to zero (1.2 percent as of June 2016 for the 16 banks composing the stress test sample). - Liquidity: loan-to-deposit ratio of 65 percent and liquid assets-to-short-term liabilities ratio of 68 percent; standard LCR metric at 80 percent in 2017, which all 16 banks meet by a comfortable margin. ### Banking sector vulnerabilities - Low risk density: risk-weighted assets amount to less than one third of overall assets in June 2016, and less than 30 percent at internationally-oriented banks; may reduce loss-absorbing capacity of capital. - Two-thirds of Luxembourg banks’ total credit exposures are treated under Basel II standardized approach. - Exposures to the public sector account for approximatively 16 percent of total assets (including ones to the Luxembourg central bank). - Exposures to group entities (internationally operating banking groups) account for about one-third of total assets. - Credit risk RWA to net loans ratio is equal to 48 percent when excluding traded assets. - Median risk weights in IRB portfolio (June 2016): Corporate Exposures 47 (Lux. banks) vs 51.8 (European banks); Retail Exposures 24 vs 27; Mortgage Exposures 17 vs 15.5. - Mortgages under IRB approach make up for less than 5 percent of total assets. - Asset quality monitoring needed: provisioning ratio relatively low at 42 percent despite negligible current NPL share; rising household indebtedness and floating lending rates could increase household vulnerability to interest rate spikes, economic downturns, or real estate price reversals. - Contagion risks: substantial integration with the euro area implies exposure via valuations of foreign assets and potential withdrawals of foreign parent funding; interconnectedness with investment funds could transmit external shocks via investor redemptions. ### Stress test design and scenarios - Horizon: both baseline and adverse macroeconomic scenarios stretch over a three-year horizon. The first year of the shock would be 2017 and the scenario runs until 2019. - Calibration variables: real GDP growth, CPI inflation rate, unemployment rate, the three-month interbank rate, the nominal government bond rate, lending rates, the euro-dollar exchange rate, and real estate price growth. - Scenarios: - Baseline: based on October 2016 World Economic Outlook projections. - Adversely severe scenario: V-shaped GDP profile producing a "triple-dip" in Luxembourg (and European) economy, driven by shocks from the Risk Assessment Matrix; projections based on IMF Global Macro-Financial Model (GMF) for external variables and a Bayesian VAR for domestic variables. Scenario includes surge in financial market volatility, financial system outflows, and renewed euro area recession. - Result: a 1.5 percent per annum decline in GDP over 2017–2019, leading to a cumulative shock of 14.7 percentage points (equivalent to a 2 standard deviation shock). ### Credit risk emphasis and exposures - Total loans represent 58 percent of total banking sector assets. - Mortgage loans comprise 4 percent of total assets but 19 percent of the assets of the five domestically-oriented banks. - Debt securities represent 23 percent of total assets, most of which are marked-to-market. - Loan book exposure extends to euro area corporates outside Luxembourg, especially Belgium, France, Germany and the Netherlands. - For nine banks under Basel II IRB, largest exposures are to other institutions (including parent banks), followed by large corporates and central governments. ### Stress test credit-loss projections and drivers - Expected credit losses over three years in adverse scenario: €2.7 billion, equivalent to 0.6 percent of total banking system assets. - Baseline scenario new provisions: €0.6 billion, equivalent to 0.1 percent of total banking system assets. - Net income before losses in the baseline: 1.8 percent of total assets (which more than offsets baseline new provisions). - PD increases: corporate SME loan portfolio PD climbs from 1.7 percent in June 2016 to 13 percent in 2018 in the adverse scenario. - Application of international benchmarks produced much more severe PD projections than national authorities’ Luxembourg-based empirical projections. - Data limitations: no allowance made for credit guarantees that Luxembourg subsidiaries may receive from their foreign parents. ### Market risk assessment (no macro hedges assumed) - Market risks included: interest, exchange rate, and equity risks. - Interest rate risk on net interest income assessed using time-to-repricing buckets (data for buckets: less than one month; 1 to 2 months; 2 to 3 months; 3 to 6 months; 6 to 12 months; and more than 12 months). - Adverse scenario interest rate path: 4-year euro swap rate projected to increase by 0.3, 0.1, and 0.1 percentage points in 2017, 2018, and 2019, respectively. - Repricing gap: five banks display a negative time-to-repricing gap; aggregate repricing gap for maturities below one year amounts to €16 million as of June 2016 for the sixteen banks. In the adverse scenario, this repricing gap and interest rate changes do not translate into a material effect on the CAR over the stress horizon. - Valuation effects on debt securities: in absence of duration data for trading portfolios, average maturity of sovereign portfolio exposure by country from EBA Transparency exercise used as proxy; losses computed as product of bond portfolio size, average maturity, and change in interest rate. *International Monetary Fund staff summary based on cr17261 (section excerpts).* ### 39. Potential valuation losses on foreign sovereign and corporate debt appear significant ### 39. Potential valuation losses on foreign sovereign and corporate debt appear significant ### Valuation losses on AFS and HFT bond portfolios - In the adverse scenario, losses due to a decline in the price of sovereign and corporate securities in the Available-for-Sale and Held-for-Trading portfolios amount to €2.5 billion. - These losses contribute by 1.4 percentage points to the decline in the CAR over the entire stress horizon. - Factors explaining the result: - Average AFS and HFT exposure of 12.5 percent of total assets for Luxembourg banks; individual bank ratios range between 0 and 41 percent. - Significant increase in vulnerable government bond rates under the adverse scenario (up to 3 percentage points for the 10-year rate), resulting in large haircuts on bond prices. - Economic hedges could not be taken into account for the other comprehensive income calculations (AFS portfolio) due to absence of appropriate data. - Mitigating factor: moderate average maturity of Luxembourgish banks' bond portfolio, averaging 4 years. *Sources for underlying estimates: Bloomberg, and IMF staff estimates.* ### Foreign Exchange Rate Risk (direct effects) - Net open FX positions were grouped by U.S. dollars and British pounds. - Banking system positive net foreign exchange position implies market gains in the case of a euro depreciation. - Net open FX position for the banking system as of June 2016: €6 million, equivalent to 0.02 percent of Tier 1 capital. - Losses on banks' net foreign exchange positions are negligible in the adverse scenario: - In the adverse scenario the euro is expected to rise against the U.S. dollar over the whole period, resulting in a small loss of €0.4 million. ### Results of the Solvency Stress Tests Based on Macro Scenarios - Under the severe adverse scenario, credit losses and market valuation losses are the main channels through which risks materialize. - Aggregate CET1 ratio (fully-loaded Basel III) would drop by 3.5 percentage points, from 18.6 percent as of June 2016, to 15.1 percent. - Contribution breakdown: - Credit losses: -1.6 percentage points of RWAs. - Market losses in the bond portfolio: -1.4 percentage points of RWAs. - Change in risk-weighted assets: -2.5 percentage points. - Offsetting factors (widening interest margin and other effects): +2 percentage points. - Banks with capital below regulatory thresholds: - Capital of a number of banks making up less than 10 percent of the banking sector’s assets would fall below the minimum regulatory Capital Adequacy ratio of 8 percent. - Recapitalization needs would be manageable, amounting to 0.8 percent of GDP. - Based on the CET1 hurdle rate of 4.5 percent, the capital shortfall would be reduced to 0.1 percent of GDP. - One more bank would need to use part of its capital conservation buffer but would remain above the regulatory minimum. - Higher vulnerability of these banks results from lower profitability, lower initial loan portfolio quality, and a relatively larger bond portfolio. - Leverage ratio effects: - Tier 1 / total assets (leverage ratio) would decline from 5.9 to 5 percent. - Six banks would see their ratios decline below the hurdle rate of 3 percent in 2019, by an amount equivalent to 1.6 percent of GDP. - Baseline scenario: - In the baseline scenario every bank would stay above the regulatory CET1 minimum. - Capital shortfalls would be larger on the basis of bank-specific Pillar II requirements (not publicly available). - Caveats and limitations: - Top-down stress test did not take into account loan write-offs and cures due to data unavailability. - Risk mitigants such as economic hedges and financial guarantees were not considered because of data issues and lack of information about enforceability during a crisis. - Some assumptions were made for Loss Given Default and other supervisory parameters due to lack of historic depth and relevant crisis experience. - Some balance sheet items, such as derivatives, were not subject to stress. - Comparison with banks' own sensitivity test (Short-Term Exercise 2015): - Banks' estimates: aggregate CET1 ratio would fall by 5 percent in a 200 bp interest rate shock after taking into account available macro/economic hedges. - IMF methodology: decline of 9 percent for the same shock. - No bank would have a CET1 ratio below 4.5 percent under banks' estimates, but three banks would need to use part of their capital conservation buffers and have a total CAR below 8 percent. - System-wide CET1 ratio: banks' estimate 15.5 percent in 2019, IMF methodology 15.1 percent. ### Policy recommendations (from stress test results) - Ensure banks' Internal Ratings-Based (IRB) models: - Take into account losses in stressed environments. - Are sufficiently conservative and do not place too high a weight on recent benign-period data. - Benchmark PDs and LGDs for retail portfolios against recent international experience for stressed macroeconomic scenarios. - Supervisory actions: - Supervisors should be able to effectively challenge banks if evidence shows aggressive modeling techniques lowering risk parameters. - Regulatory measures: - Commend and finalize shortly the European authorities' plan to introduce regulatory floors for the risk-weights calculated by IRB banks. ### Market Risks: Sensitivity Analyses and Key Test Assumptions - Sensitivity tests are static (instantaneous impact as of June 2016) and assume constant risk-weighted assets after shocks. - Table of comparison between severely adverse scenario projections and sensitivity test assumptions (as used in the sensitivity tests): - Lending rate shock: Severely adverse scenario 86 bps; Sensitivity test 500 bps. - Domestic sovereign interest rate shock: Severely adverse scenario 177 bps; Sensitivity test 500 bps. - Foreign exchange rate shock: Severely adverse scenario -7.4 percent; Sensitivity test +/- 30 percent. - Equity price shock: Severely adverse scenario -19 percent; Sensitivity test -50 percent. ### Sensitivity test results: sovereign, interest rate, FX, equities, concentration - Domestic sovereign securities (500 bp interest rate increase on AFS and HFT portfolios): - CAR would decline by 0.3 percentage points due to these losses taken in isolation. - Luxembourgish banks are little exposed to domestic sovereign bond risks; initial capitalization would avoid undercapitalization from this shock. - Interest rate risk (500 bp increase, maturity ladder/gap analysis): - The shock taken in isolation would lower total banking system's net interest income by €0.8 million. - No impact on aggregate Tier 1 capital ratio; would not cause undercapitalization in any of the sixteen banks. - Foreign exchange rate sensitivity (30 percent appreciation/depreciation of euro nominal effective exchange rate): - 30 percent appreciation of the euro would cause a loss of €1.8 million; very limited impact on aggregate Tier 1 ratio; would not cause undercapitalization in any of the sixteen banks. - 30 percent depreciation of the euro would cause a gain of €1.8 million. - Equity price shock (50 percent decline): - Marked-to-market equity investments average 0.8 percent of total assets, range 0 to 2.2 percent. - A 50 percent decline in stock prices would cause a loss of €1.75 billion, translating into a 1.3 percentage point decline in the aggregate Tier 1 ratio. - This shock, taken in isolation, would not cause undercapitalization in any of the sixteen banks. - Concentration risk: simultaneous default of large corporate exposures - Average single gross largest exposure: 22.4 percent of Tier 1 capital. - Average net largest exposure (after cash collateral and credit risk mitigation): 8.8 percent of Tier 1 capital. - Default of the net largest exposure of each of the sixteen banks would not cause undercapitalization. - Simultaneous default of the five net largest exposures: - Would lead three banks to be undercapitalized with regard to the Tier 1 capital ratio of 6 percent. - Translates into a capital shortfall of 1.6 percent of GDP. - Simultaneous default of the ten net largest exposures: - Would cause six banks to be undercapitalized in terms of Tier 1 capital. - Raises the capital shortfall to 8.9 percent of GDP. - Additional tests: - A 30 percent additional haircut on real estate collateral (on top of regulatory haircut) does not change the number of undercapitalized banks or capital shortfalls in these scenarios (reflecting that banks do not hold most credit risk mitigation as real estate). - If exempted corporate exposures are reintegrated and a zero recovery rate is assumed on credit risk mitigation measures: - Number of undercapitalized banks following default of the largest, five largest and ten largest exposures would increase to one, six and eight respectively. - Capital shortfalls would increase to 1.2 percent, 11.6 percent and 24.4 percent of GDP respectively. - These scenarios assume a "double default" (borrower and financial guarantor) and no credit risk mitigant; they are based on extremely severe assumptions. *Italic: IMF staff summary based on chapter "Potential valuation losses on foreign sovereign and corporate debt appear significant" (cr17261).* ### 59. The household stress test was used to benchmark the banks' credit loss projections in ### cr17261 - 59. The household stress test was used to benchmark the banks' credit loss projections in ### Household stress test: methodology and data - A household stress test based on micro data and household characteristics was performed to complement banks' internal models and assess solvency risk and non-linear effects of shocks. - Data source: 2014 micro data on Luxembourgish households from the Eurosystem Household Finance and Consumption Survey (HFCS). - Methodological note: The description of the methodology is provided in Appendix VI. The implied PDs from the household stress test are not realized or true forward-looking PDs as those used for the banking sector solvency stress test (see Appendix V for details). ### Household stress test: scenario applied - Scenario applied: the same severely adverse macroeconomic scenario as for the banking sector stress test, including: - a cumulative drop of 30 percent in house prices over three years, - a 5 percentage point rise in the unemployment rate, - an annual drop in wages and unemployment benefits by 5 percent. ### Household stress test: key results and statistics - Implied households' PDs under the shock: - current implied PD: 2.5 percent (as estimated using micro data), - peak implied PD under the scenario: 6.4percent (reflecting a multiplier of 2.6). - Liquidity and repayment buffers: - more than 93 percent of indebted households have sufficient liquid assets to cover at least 12 months of income shortages (debt service payments and basic living costs larger than current income). - Distributional note: - share of outstanding balance of mortgage debt for the 1st decile (households with highest PD) is equal to 4.1 percent of total outstanding balance of mortgage debt. - Comparative assessment: - The stress test provides implicit probabilities of default in line with projections made within the banking sector stress test framework and with the Central Bank of Luxembourg’s own estimates. - Interpretation: - Households' solvency would be significantly impacted by the combined shocks, but the proportion of households in default would remain manageable. --- ### Solvency analysis and stress test of the investment fund sector: scope and sample - Emphasis on UCITS funds which account for 83 percent of AuM in Luxembourg (€2,847 billion as of March 2016); Alternative Investment Funds AuM: €445 billion. - Stress test sample: - 191 funds (all UCITS except one) amounting to €656 billion, - Coverage: 75 percent of the AuM of HY and EM bond funds in Luxembourg, 54 percent of mixed funds investing primarily in fixed income instruments, 32 percent of the large bond funds and 75 percent of the Money Market Fund industry in Luxembourg. - Aggregate totals (as presented in the sample): - Total net assets in sample: €656 billion, - TNA universe: €1,336 billion, - Total coverage: 49.1 percent (of TNA universe). ### Investment funds: concentration risk and common exposures — findings - Bond and mixed fund exposures: - Portfolios characterized by concentrated holdings in sovereign issuers; UCITS limit of 5 percent per issuer is waived for sovereigns. - Around half of Luxembourg's EM bond funds in the sample have exposures to single EM sovereign issuers in excess of 5 percent of TNA, with exposures for a number of funds up to 30 percent of TNA. - EM bond funds have significant common exposures (holdings of Brazil, Mexico, Turkey and Indonesia amounting to more than 30 percent of TNA). - Large parts of bond and mixed funds (accounting for respectively 34 and 46 percent of TNA in the sample) also have large exposures to individual sovereigns, typically developed country issuers (U.S., Germany, Italy, and France). - HY bond funds: - Limited concentration risk and low common exposures due to 5 percent issuer limit under UCITS, but exposed to low-quality issuers. - MMFs: - Among the sample of 27 MMFs, only 4 have exposures to individual sovereigns higher than 20 percent and they are all towards developed countries (U.S., Italy, and Spain). ### Authorities' monitoring and recommendation on funds - Current monitoring: - The CSSF and the BCL do not currently monitor concentration risk and common exposures across funds on a regular or systematic basis. - Both authorities have access to monthly security-by-security data for all Luxembourg-domiciled funds but use the data for statistical purposes (BCL) or on an ad hoc basis (CSSF). - Recommendation for CSSF: - Intensify monitoring of concentration risk and common exposures within the fund industry, leveraging the UCITS risk reporting template and further data gathering. - Obtain access to the ECB’s Centralized Securities Database to fill possible data gaps. - Ensure funds with high concentration risk have a specific and adequate risk management process in place, including tools to mitigate liquidity and market risks (e.g., access to a set of Liquidity Management Tools to be used in stress). --- ### Solvency stress test for Money Market Funds (MMFs) - CNAV MMFs: - Offer ability to investors to redeem at par due to amortized cost accounting; invest in short-term fixed income instruments valued at amortized cost rather than mark-to-market. - Investors may view the 'redeemable at par' feature as a substitute for bank deposits despite market and liquidity risk. - Historical note: - During the GFC one CNAV MMF in the U.S. ‘broke the buck’ with the shadow NAV more than 50 basis points lower than par, leading to large redemptions. In Luxembourg, the largest deviation between shadow NAV and par for the 10 largest CNAV MMFs remained below 20 basis points except for one MMF slightly above 20 basis points; no CNAVs suspended redemptions or used gates. - Stress test sensitivity results: - A combination of a 170 bps shock to risk free interest rates and an increase in credit spreads on non-sovereign holdings of 70 bps would be required for the shadow NAV to deviate substantially (by 0.2 percent or more) from par value (asset-weighted average of five largest CNAV MMFs). - Solvency risks are mitigated by CNAV MMFs holding large cash buffers and high credit quality short-duration assets. - For the five CNAV MMFs in the sample: - deposits and reverse repo amount to almost one quarter of portfolio holdings, - short-term debt instruments account for more than half of portfolio holdings (most comprised of the highest short-term credit rating). - Notes on thresholds: - The 0.2 percent threshold is used in industry guidelines and some national/regulatory transpositions; there is no specific threshold in the UCITS Directive. ### Recommendations on MMFs - CSSF should, taking into account the requirements of the upcoming new-MMF Regulation: - provide guidance on market risk-based stress tests for MMFs using amortized cost, and possibly for LVNAVs in the upcoming EU framework; - provide asset managers with broad guidance on how to perform solvency stress tests for CNAVs, covering modalities including coverage, frequency and scenario design; - ensure stress tests cover current CNAVs and the new types of MMFs contemplated under the EU proposal (Public Debt CNAV and Low-Volatility Net Asset Value MMFs); - build internal capacity for performing stress tests to extend eventually to the broader investment fund industry. --- ### Insurance sector: solvency stress test (summary) - Scope and coverage: - Bottom-up solvency stress tests coordinated by the CAA (limited IMF input); coverage included 10 life insurers (71 percent of technical provisions) and the one dominant reinsurer (accounting for two thirds of sector assets). - Share of unit-linked products in assets of the 10 life insurers in the sample: 77 percent. - Scenario features: - FSAP scenario featured a "double hit" with interest rate and equity price shocks; differed from 2016 EIOPA stress test in market shock size, embedding of companies' business plans in projections, and allowing companies to recalculate solvency capital requirement in the adverse scenario. - Results: - Aggregate Eligible Own Funds decline by 27 percent in the third year of the scenario compared to the 2016 baseline. - Each of the 10 life insurers, plus the reinsurer, maintained a Solvency II capital requirement well above the minimum requirement of 100 percent. - Scope note: - Non-life sector not included in stress tests due to idiosyncratic nature of shocks; non-life insurers comprise the smallest segment of the insurance industry. --- ### Liquidity stress tests (overview) - Banking sector: - TD liquidity stress tests assessed banks' capacity to withstand large withdrawals by assuming various bank-run scenarios and differentiating flows by maturities and currencies. - Investment funds: - Investment fund liquidity stress tests assessed ability to meet redemption shocks calibrated on historical net flows and estimates derived from the macroeconomic adverse scenario used for the banking sector stress tests. *Source: IMF staff report excerpt (Luxembourg).* ### 73. Three types of liquidity stress tests were performed in order to arrive at a holistic ### Three types of liquidity stress tests were performed in order to arrive at a holistic assessment of bank liquidity risks. ### Types of stress tests (overview) - Three complementary liquidity stress tests were performed: - Liquidity Coverage Ratio (LCR) based test: measures ability to meet liquidity needs in a 30-day stress scenario using a stock of unencumbered high-quality liquid assets (HQLA). Phase-in: 60 percent in 2015, 70 percent in 2016, converge to 100 percent by January 2018. - Net Stable Funding Ratio (NSFR) based test: provides a complementary view of banks' funding profile at a one-year horizon (effective from January 2018). - Cash flow (outflow) analysis by maturity buckets: granular analysis of liquidity buffers across maturities ranging from 1–7 days to more than a year. - Data coverage: - LCR results: supervisory information as of September 2016 covering all 16 banks in the sample. - NSFR and cash flow analysis: supervisory information from Short Term Exercise (STE) templates as of June 2016, covering only half the sample due to data availability. ### LCR-based stress test — scenarios and assumptions - Three aggregate-currency scenarios (also applied separately to major foreign currencies EUR, USD, GBP under standard assumptions): - Standard LCR scenario: parameters as set out by Basel III LCR 2013. - Luxembourg retail stress scenario: deposit run assumptions — run-off rates raised to 20 percent for stable and 30 percent for unstable retail and wholesale deposits. - Luxembourg wholesale stress scenario: calibrated to link investment fund stress with interbank market stress. Key assumptions include: - run-off rates of up to 100 percent for wholesale funding from other legal entity customers; - rates of 50 percent for operational deposits generated by clearing, custody, and investment fund activities; - outflows by non-financial corporates, central banks and multilateral development banks of up to 70 percent. - Table 6 (excerpted assumptions) includes exact run-off and roll-off rates and haircuts used across outflows, secured funding, derivatives, ABCP/SIVs, undrawn facilities, and other contingent funding liabilities. ### LCR-based stress test — key results - Standard scenario (aggregate LCR as of September 2016): - Aggregate LCR = 109.4 percent. - Four out of 16 banks do not pass the 100 percent hurdle rate (binding in 2018). - Banks meet prevailing 80 percent hurdle rate for 2017. - Shortfall at 100 percent hurdle rate = €3.5 billion or 0.8 percent of banks' assets. - Luxembourg retail stress scenario (20–30 percent retail run-offs): - Aggregate LCR would fall to 107.6 percent. - Four out of 16 banks would fall below 100 percent; three would continue to meet 80 percent. - Liquidity shortfall = €1.9 billion or 0.4 percent of banks' assets. - Context: severity exceeds global financial crisis peak outflow rates (8.6 percent household, 22.4 percent non-financial corporate). - Luxembourg wholesale stress scenario (50–100 percent unsecured wholesale run-offs): - Aggregate LCR would fall to 96.3 percent. - Seven banks below 100 percent; two below 80 percent. - Private banks LCR falls to 85 percent; domestically oriented banks to 92 percent. - Total liquidity shortfall = €7.2 billion or 1.6 percent of banks' assets. - Scenario noted as extreme but informative for bank–investment fund linkages. - Foreign-currency LCR tests (EUR, USD, GBP using standard scenario assumptions): - Liquidity shortfalls range between €0.3 billion and €12.7 billion, i.e. 0.1 percent and 3.5 percent of banks' assets. - Particular weaknesses in U.S. dollar positions due to foreign subsidiary structure and reliance on parent groups for FX management. ### NSFR-based stress test — key results - Aggregate NSFR (June 2016): - Available stable funding = €165 billion. - Required stable funding = €160 billion. - Aggregate NSFR = 101.1 percent (slightly above 100 percent minimum). - Most banks meet the hurdle with significant variation by business model. - Two banks fall short; shortfall larger than under LCR wholesale shock scenario. - Authorities confident those two banks are likely to be fully compliant with NSFR by January 2018. ### Outflow analysis (cash flow based) — methodology and results - Maturity buckets used: 1 to 7 days; 8 days to 2 weeks; >2 weeks up to 4 weeks; >4 weeks up to 3 months; >3 months up to 6 months; >6 months up to a year; more than a year. - Analysis captured: - (i) liquidity needs from outflows; - (ii) available standby liquidity from inflows; and - (iii) buffers to counterbalance liquidity gaps. - Counterbalancing capacity compared net outflows to liquid assets available for sale, with liquid assets subject to 20 percent haircuts. - Key findings: - All banks except one would be able to meet substantial funding gaps in 1 to 7 days by selling liquid assets equal to 9.2 percent of bank's total assets. - Analysis limited to eight banks (STE coverage), excluding the largest bank, which restricts full-system liquidity analysis. - Reliance on securities to fund short-term gaps is a vulnerability if many banks try to sell assets simultaneously at prices below par. ### Liquidity risks and vulnerabilities (synthesis) - Tests confirm strong liquidity buffers in Luxembourg banking system but highlight short-term liquidity exposures: - Reliance on funding from other financial institutions (intragroup and wholesale funding). - Foreign currency funding and FX management reliance on parent groups. - Reporting and data gaps: limited coverage for NSFR and outflow analysis (STE covers only eight banks). ### Policy recommendations and supervisory implications - Require banks to lengthen the maturity of unsecured wholesale funding beyond a seven-day window, including funding from parent groups. - Improve liquidity monitoring by performing liquidity stress tests using cash flow structures at various maturities and by currency (FX). - Consideration for ECB: implement a binding FX LCR framework at the group level to address difficulty verifying LCR in FX at group level frequently. - Extend coverage of liquidity reporting to more banks; authorities should work with the EBA to close liquidity reporting gaps and expand harmonized EU bank reporting. ### Liquidity risks in the investment fund industry — methodology (summary) - Fund liquidity stress test assesses ability to withstand severe redemption shock with minimal disruption, comparing liquid assets to net outflows under two approaches: - Historical approach: redemption shock = 1 percent highest monthly net flows observed over 2007–2016 for each fund (each fund faces a different shock). - Forward-looking (macro) approach: econometric model relates net flows at strategy level (EM, HY, mixed, other bond funds) to macrofinancial variables; redemption shock calibrated using values under the adverse macro scenario (same shock across funds within a strategy). - Measures of liquid assets: - (i) cash and short-term debt securities (residual maturity < one year); - (ii) HQLA approach with liquidity weights by credit rating, instrument type and issuer (per ESMA 2015). - Redemption Coverage Ratio (RCR) defined as: RCR = Liquid assets / Net outflows. RCR < 1 implies insufficient liquid assets and need to sell less liquid assets and/or use Liquidity Management Tools (LMTs). ### Investment fund liquidity test — key quantitative findings - Redemption shocks (percent of total net assets, TNA): - Historical approach (average net flows): EM 18%, HY 19%, Mixed 9%*, Other bond funds 18%, Short Term CNAV 19%**, Short Term VNAV 23%**, Other MMFs 18%**. - Macro model (net flows): EM 9%, HY 11%, Mixed (net inflows) 11%*, Other bond funds 6%. (MMFs excluded from forward-looking approach.) - Notes: *Under the adverse scenario, mixed funds would experience net inflows of 11%. **The model is not significant. - Historical approach results: - MMFs: liquidity buffers measured by short-term debt securities and cash would cover simulated redemption shocks. - Mixed, EM and other bond funds: short-term assets might not be enough; under HQLA approach only a few funds would have RCR < 1 due to sovereign exposures. - HY bond funds: lower RCR under both measures; would need to sell less liquid assets (fire sale risk), use LMTs, or borrow from banks. - Fund manager self-assessments (UCITS reporting) indicate more benign liquidity risks, based on low number of days to sell portfolios, contrasting with stress-test metrics. - Additional note: - Non-investment grade corporate bonds are given a weight of zero in the HQLA measure, resulting in relatively low HQLA for HY funds; HQLA measure does not account for possible liquidation of non-investment grade bonds. *Sources: CSSF, BCL, and IMF staff calculations.* ### 87. Under the historical approach, the estimated liquidity shortfall for some funds would ### cr17261 - 87. Under the historical approach, the estimated liquidity shortfall for some funds would ### Historical approach: liquidity shortfalls under UCITS - The liquidity shortfall is defined as the difference between the redemption shock and liquid assets (HQLA measure) expressed in percent of TNA. - For some HY funds and a few bond funds, the shortfall would be above 10 percent of total net assets, the maximum (temporary) borrowing limit under the UCITS Directive. - Empirical/data notes: - The data refer to March 2016 and is based on the half-yearly CSSF UCITS Risk report. - The report has been recently launched and therefore data quality issues might impact the interpretation of the data. - Example: CNAVs reported less liquid assets (up to one day) than HY funds, due to different interpretations by asset managers, particularly by one CNAV where there was a misinterpretation on final maturity and time to liquidate, but where the fund was very liquid. - Data reported refer to liquidity under normal conditions; reporting on liquidity under stress was optional for the first report and is not reported here due to low coverage. ### Forward-Looking Approach: stress-test outcomes under a macroeconomic scenario - Overall stress-test results are milder under the macroeconomic scenario versus the historical approach. - Key findings: - Most HY funds would need to liquidate less liquid security holdings to cope with redemptions or use liquidity management tools. - All bond funds except HY funds would have enough HQLA to cover the redemption shock. - Around 70 percent of HY funds would experience liquidity shortfalls (a slightly lower share than in the historical approach). - When liquidity is measured by short-term assets, a significant share of EM and other bond funds would have an RCR below one, though the proportion would be lower than under the historical approach. - Table/figure references: Table 11 and Figure 19 illustrate detailed results; Table 10 presents Luxembourg: Liquidity Stress Test Results for the Forward-looking Approach. - Sample statistics and coverage noted in the source: - Coverage: 75 percent of the total net assets for HY, EM bond funds, VNAVs, and MMFs; 54 percent for mixed funds; and 32 percent for bond funds. ### Potential bank deposit drawdowns implied by stress tests - Modeling results indicate that, absent adequate liquid securities and effective LMTs, investment funds might need to draw down on bank deposits. - Estimated deposit outflows: - Around 25 percent of total cash deposits of investment funds in the sample, equivalent to around €5.2 billion. - Close to 20 percent (€4.9 billion) when incorporating projected inflows into mixed funds. - Data caveat: - Due to data gaps, deposit outflows are estimated assuming that 100 percent of funds deposits are with their depositary bank, while in practice funds’ deposits are likely to be spread across several banks domiciled inside and outside Luxembourg. ### Work done by the authorities (CSSF and BCL) - CSSF monitoring and reporting: - The CSSF monitors fund liquidity risk to ensure compliance with UCITS regulations but does not require regular liquidity stress tests for funds. - Under UCITS, liquidity stress tests are required "where appropriate"; CSSF Regulation transposing UCITS does not provide guidance on liquidity stress tests. - Ongoing supervision includes the requirement for a risk management process, including liquidity risk management and stress testing. - The CSSF has asked, in certain cases, Management Companies to provide further data and to improve their liquidity stress tests policies. - Many Management Companies perform liquidity stress tests on a regular and systematic basis despite not always being required to do so. - Reporting initiatives: - The half-yearly UCITS Risk Reporting requests a wide range of data, including portfolio liquidity by buckets under normal and stress periods; the second UCITS risk report was in December 2016. - The CSSF also asks for quarterly reporting for UCITS using leverage and has used ad hoc data collection to analyze risks linked to securities financing transactions. ### Recommendations (policy actions for CSSF and BCL) - Liquidity stress-test guidance and requirements: - The CSSF should provide industry guidance on liquidity stress tests for funds most exposed to less liquid asset classes. - The CSSF should require such funds to perform liquidity stress tests on a regular basis, ideally using a scenario provided by the Authorities to ensure comparability across funds. - The CSSF should review and challenge such liquidity stress tests, which will require building internal capacity. - Liquidity management tools (LMTs) assessment: - The CSSF should ensure funds exposed to potential liquidity mismatches have adequate LMTs in place and should assess their effectiveness. - For funds that could have significant liquidity shortfalls under stress, an adequate set of LMTs (e.g., suspension of redemption, redemption gates, swing pricing) should be in place. - The CSSF should, in line with international developments, commission a study on the effectiveness of LMTs to provide the basis for subsequent industry guidance. - Bank–fund interlinkages and data collection: - The CSSF and the BCL should continue to analyze risks arising from bank-fund interlinkages. - Data gaps should be closed by collecting data on individual funds' exposures to banks in Luxembourg to assess depositary banks' vulnerabilities to stress in the fund industry. ### Interconnectedness analysis and contagion risks — methodology and coverage - Purpose: Assess systemic risks by investigating interlinkages within and across sectors to uncover potential contagion and amplification channels. - Methods referenced: Espinosa-Vega and Solé (2010), Segoviano and Goodhart (2009), Diebold and Yilmaz (2014). - Data coverage notes: - Market-based network analysis for banks covered only four banks due to data availability; therefore general conclusions from that analysis are limited. - Investment fund coverage was extensive: 75 percent of TNA for HY, EM bond funds, VNAVs, and MMFs; 54 percent for mixed funds; and 32 percent for bond funds. ### A. Domestic interconnectedness analysis — main findings - Interbank network: - Domestic interbank positions are small compared to banks' capitalization. - For each of the sixteen banks in the sample, total gross exposures to the other fifteen banks are smaller than its regulatory capital. - No single failure of a domestic bank would trigger the failure of another bank in the sample; aggregate indices of contagion and vulnerability appear low. - Example indices: - The entity with the highest contagion index causes losses of less than 0.4 percent of counterparties' capital. - The entity most adversely impacted incurs losses of less than 1.5 percent of its capital. - Market-data-based network analysis confirms limited interconnections between domestic banks, except for two banks. - Connectedness has declined since the European debt crisis; the JPoD and the Diebold-Yilmaz total connectedness indicator both suggest declining risks and connectivity, with connectivity declining in 2016. - Bank–insurer network: - Contagion stress test based on supervisory data covering 16 banks and 12 largest Luxembourgish insurers (including capital cross-participation, bonds, credit and deposits). - Limited potential for spillovers within and across banking and insurance sectors. - Aggregate indices low: - The entity with the highest contagion index causes losses of less than 0.5 percent of counterparties' capital. - The entity with the highest vulnerability index incurs losses of less than 2.5 percent of its capital. - Banks tend to have higher indices of contagion; insurers tend to have higher indices of vulnerability. - All bank exposures to insurers are through debt security holdings; the vast majority of insurer exposures to banks are in the form of deposits; no direct cross exposures amongst insurers. - Domestic interfund and bank–fund networks: - Market-data network analysis shows strong connections between some investment funds and with two banks. - Bond, mixed and HY investment funds are closely related to each other; the bond node appears most interconnected. - MMF, EM and VNAV funds are located more externally in the network and are not strongly connected to other entities. - Two banks (Bank 3 and Bank 1) appear closely related to bond and mixed funds but relatively independent from other entities in the network. ### B. Cross-border contagion risks - Cross-border balance sheet analysis shows strong linkages with euro area members and countries with deep financial sectors. - Cross-border exposures between subsidiaries and their foreign parents in the euro area are particularly strong. - Significant exposures exist outside the euro zone, notably with the UK, the US and Switzerland. - Network visualization indicates Luxembourg banks occupy an inner core with strong interconnectedness to foreign banks (outer circle), especially foreign parents. _Italic: Sources: CSSF, BCL, and IMF staff calculations._ ### 105. Germany is the largest recipient of Luxembourg-domiciled banks' assets as well as the ### 105. Germany is the largest recipient of Luxembourg-domiciled banks' assets as well as the ### Cross-border counterparties and activities - Germany is the largest recipient of Luxembourg-domiciled banks' assets and the largest source of funding to Luxembourg banks. - More than half of Germany's shares are associated with corporate finance activities. - France, the UK and the US are the next three largest recipients and: - account for one-third of total foreign assets, - provide a combined funding of only less than half of that of Luxembourg banks. - A significant share of exposures to France, the UK and the US are associated with custodian banking activities; for France, private banking and domestically-oriented activities also play an important role. - Switzerland is nearly balanced between being a recipient and a funding source, where mostly private banking activities make up its share. - Cayman Islands and Singapore appear in the top 10 as important sources of funding. ### Asset and liability composition by transaction type (as of 2016Q2) - Assets: claims in the form of loans to affiliated entities make up the largest share of cross-border exposures. - Liabilities: debt in the form of overnight deposits, deposits with agreed maturity, deposits redeemable on notice and repos and short sale securities constitute almost three quarters of foreign sources of funding. - More than half of this debt (equivalent to about 40 percent of total foreign liabilities) is owed to affiliated entities. - Chart-level shares reported (as displayed in source figures): - Assets composition values shown: 39%, 30%, 27%, 1%, 2%, 1%, 4% (as labeled in the source figure). - Liabilities composition values shown: 25%, 28%, 14%, 4%, 10%, 13%, 4%, 2%, 19% (as labeled in the source figure). ### Intragroup balance sheet interconnectedness - Balance sheet analysis confirms Luxembourg entities are highly exposed to their foreign parent groups, almost entirely via loans and advances. - Weighted average exposure across banks is almost 6 times the subsidiary's capital (compared to a maximum of 25 percent without the waiver). - Weighted average exposure from parents in funding terms is around 2.5 times liquid assets of the subsidiary. - Custodian banks have uniformly higher exposures to intragroup entities (result of upstreaming deposits). - Banks focused on intragroup liquidity management have uniformly higher share of intragroup funding. - Sample and data notes: - Analysis relies on FINREP Related Parties template reported at the highest level of consolidation for each bank in the 16-bank sample. ### Network model and data sources used - Main analytical basis: Espinosa-Vega and Sole (2000) model applied to supervisory data. - Primary data source: COREP large exposure template (breakdown of each bank's assets by counterparty). - Complementary data: COREP template on Concentration of Funding by Counterparty. - Contagion risk assessment relied on supervisory data on banks' large exposures and funding sources provided in a secure room at the ECB. - Scope of datasets after aggregation and filtering: - (i) the 20 Luxembourgish reporting banks; - (ii) 8 additional smaller banks in Luxembourg (not part of the sample); - (iii) 26 parent holding groups of the Luxembourgish banks; - (iv) 104 other banks. - Exposures vis-à-vis clients amounting to less than 100,000 euros were filtered out. - Data focused on gross original exposures; credit risk mitigation and collateral were not taken into account (noting this leads to an overestimation of linkages and vulnerabilities). - Sensitivity testing performed using a wide range of loss-given-default parameters (λ = {0.05, 0.1, 0.2, 0.4, 0.6, 0.8, 1}). ### Stylized contagion model parameters and mechanics (baseline) - Baseline parameters: - λ = 0.4 (loss-given-default = 40 percent), - ρ = 0.35 (fraction of funding lost = 35 percent), - δ = 1 (asset fire-sale discount such that assets trade at a 50 percent discount in the description: “with a 50 percent discount rate on the assets that a bank may be forced to sell”). - Credit shock condition: bank i fails if k_i – λ x_hi < 0, where x_hi are bank i loans to bank h and k_i is i’s capital. - Funding shock condition: bank i fails if k_i – (λ x_hi + δ ρ x_ih) < 0, where x_ih stands for bank i borrowing from bank h. - Model outputs for each bank: - Induced failures (number of following failures if bank i fails first), - Vulnerability level (number of bank i’s failures due to other banks failing), - Index of contagion (average percentage loss of other banks due to failure of bank i), - Index of vulnerability (average percentage loss of bank i due to failure of all other banks). ### Sample-level exposures and funding (aggregate figures) - Combined sum of large exposures for the sample banks: 362 billion euros. - Total assets of the same banks: around 521 euros (note: original text states “around 521 euros in total assets of the same banks” — this appears verbatim in source). - Combined funding received from 10 largest counterparties: 220 billion euros. - Of which 162 billion euros are associated with credit institutions. - About 140 billion euros of bank-to-bank funding is related to intra-group entities. - 16 billion euros in unsecured wholesale funding from financial counterparties. - Combined total liabilities referenced: a combined 484 billion euros in total liabilities. ### Simulation results and vulnerability findings - Under gross original exposures before credit risk mitigation and baseline model assumptions: - Six banks suffer losses in excess of their capital even with a low loss-given-default assumption of 5 percent. - Seven additional banks face similar distress under higher but reasonable losses (loss-given-default range of up to 40 percent). - Contagion may also derive from a non-parent foreign bank; one custodian bank (L3) stands out as strongly connected to a number of foreign banks. - Vulnerability and contagion indices: - Two Luxembourg banks average above 25 percent losses to their capital across independently triggered failures of all parents. - Sources of contagion among parents are highly concentrated on a couple of institutions, both causing losses to a few counterparties; losses on these few counterparties represent more than 100 percent of their capital. - Example attribution: two-thirds of L3's losses can be attributed to credit shock; P14 induces losses to Luxembourg banks through credit and funding channels almost by equal amounts. - Parameters used in illustrative figure panels: Λ = 40 percent (loss given default); ρ = 35 percent (loss given funding lost). Calculations based on gross exposure, including off-balance sheet contingent liabilities and excluding cash and other collateral received and without other credit risk mitigation measures. ### Market-based network analysis (2011–April 2016) - Market-based network analysis indicates strong interconnectivity across Luxembourg investment funds and global banks, particularly for mixed, bond and HY funds which show strong connections to non-domestically located US and UK banks. - Mixed, bond and HY funds appear central in the network (node color red/orange and central location). - Luxembourg EM and MMF funds appear relatively independent from the network (outside location), a pattern likely reflecting better return performance of EM funds during the sovereign debt crisis over the sample period. - Methodological notes: - Edge thickness in network visualizations shows strength of pairwise relationship measured by 12-month forecast error variance of entity i due to shocks from entity j. - Probabilities of distress (PoD) derived from bond spreads for Luxembourg banks (bond spreads using the euro area risk free rate) and from returns for investment funds (marked-to-market assets data); CDS used for global banks when available. - Node size indicates total asset size; node color indicates total connectedness (sum of pairwise connections); node location derived using ForceAtlas2 algorithm. *Source: IMF staff analysis in the referenced chapter.* ### Box 5. Technical Summary of Market Based Models (continued) ### Box 5. Technical Summary of Market Based Models (continued) ### Banks’ Bond Yield Information and Comparison of CDS and Bond Spreads - Sources: Thomson Reuters Datastream, Bloomberg, and IMF staff calculations. - Balance sheet information from the authorities enabled a new approach where PoDs are derived by connecting redemption risk to investment funds return performance. - Luxembourg authorities provided fund-by-fund data for classes of investment funds: bonds, emerging markets (EM), high yields (HY), mixed funds (MF), variable net asset value (VNAV), constant net asset value (CNAV) and money market funds (MMF). - PoDs for banks can be derived from bond spreads (see Hull et al. 2005) using the risk-free rate, divided by 0.6 to get the risk-neutral rate (for banks), credit default spreads for global banks and balance sheet information (for investment funds). ### Derivation of Probability of Distress (PoD) for Investment Funds - Returns estimation: - Returns estimated using monthly mark-to-market asset data (i.e. total net assets; TNA). - TNA’s adjusted by monthly redemptions and subscriptions to receive the “pure” market return caused only by changes in prices. - Over a 12-month rolling window, the number of times returns fell under a specific threshold were summed-up and divided by the number of observations to get the probability of distress. - Threshold selection: - PoDs were derived using several thresholds, starting with the mean of the distribution of overall returns (not the rolling window). - PoDs were most sensible when using the 5th percentile of the overall distribution of returns as a threshold (comparable to the percentile threshold used in Basel III regulation). - For VNAVs, the 5th percentile was deemed too low (likely due to the low number of funds in this sample; only seven). Thus, the 8th percentile of the overall distribution was used for VNAVs. - Empirical performance: - The PoDs derived reflect increased risk during the global financial crisis and European debt crisis properly. - Note: - CNAVs could not be included into the exercise as the number of data points was too low. ### Combination of PoDs and Model Integration - PoDs of banks and investment funds were combined to analyze possible interlinkages. - Since PoDs are exogenous to Segoviano’s model, they can be derived from different sources, allowing inclusion of PoDs derived from bond spreads and balance sheet information. - Network analysis reveals interconnectivity between investment funds and domestic and global banks. ### Conclusion: System Resilience and Key Vulnerabilities - Aggregate resilience: - High starting levels of system-wide capital allow most banks to absorb a large shock under the adverse scenario and retain substantial buffers. - Bank liquidity displays broad resilience, but would be weakened should wholesale funding dry up. - Solvency stress tests for insurers indicate high starting levels of capital allow life insurers to absorb large market shocks while maintaining adequate capital buffers. - Liquidity and solvency risk for money market funds appears muted. - Liquidity and concentration risk suggests continued monitoring for select bond funds. - Interconnectedness: - Luxembourg banks are not interconnected at the domestic level, but quantitative analysis confirms significant intra-group exposures at the cross-border level. - Network analysis confirms interconnectivity between investment funds and domestic and global banks. ### Policy Recommendations: Banks and Liquidity - Strengthen liquidity monitoring by performing liquidity stress tests. - Extend the maturity of wholesale funding. - Implement the FX LCR at the group level. - Work together with the EBA to close liquidity reporting gaps and expand the harmonized EU bank reporting. ### Policy Recommendations: Investment Funds - Provide industry guidance on liquidity stress tests for funds investing in less liquid asset classes (including on frequency and possibly the use of a common scenario/approach to ensure comparability across funds). - Enhance monitoring of concentration risk. - Assess the effectiveness of liquidity management tools. - Further analyze vulnerabilities posed by interlinkages between investment funds and banks. ### Policy Recommendations: Contagion and Data Gaps - The ECB should work closely with other European countries and agencies on closing data gaps on cross-border bilateral exposures. - Current data limitations: - Data on bank-level cross-border asset positions vis-à-vis counterparties are available to the authorities. - Information on bilateral cross-border liability positions is still limited to the ten largest funding sources. - Maturity information on large exposures is limited for short-term exposures. - Future data collection enhancements will help monitor systemic risk related to cross-border and cross-sectoral linkages. ### Risk Assessment Matrix (RAM): Selected Entries - Sharp rise in financial market risk premia and renewed Euro area recession: - Relative likelihood: Medium - Impact if realized: High - Transmission channels: Cross-border financial intermediation contraction; trapped liquidity upstreamed to parents; synchronized redemption shock causing fire sales and drawdown of local bank deposits; operational stresses at Clearstream (CBL). - Protracted uncertainty associated with political fragmentation and protectionism: - Relative likelihood: Medium - Impact if realized: Medium - Transmission channels: Reduced trade and financial flows; potential loss of London passporting affecting UK-based banks and investment funds; Luxembourg funds’ large exposures to US capital markets. - Downturn in Luxembourg’s residential housing market: - Relative likelihood: Medium - Impact if realized: Medium - Transmission channels: Domestically-oriented banks exposed to non-performing mortgages and capital hits; household net worth decline affecting consumption; potential fiscal contingent liabilities. - Structural outflows from the financial system and operational challenges adapting to regulatory and technological changes: - Relative likelihood: Medium - Impact if realized: High - Transmission channels: Shrinkage in financial system assets; erosion of tax base (one quarter of GDP directly generated by the financial sector). ### Stress Test Matrix (STEM) — Banking Sector: Solvency, Liquidity, Contagion - Banking Sector: Solvency Risk (Top-down by FSAP Team) - Institutional perimeter: - Institutions included: 16 banks - Market share: 73 percent of the banking sector’s assets; 94 percent of residential mortgage loans; 68 percent of exposures to foreign banks - Data and baseline date: Publically-available and supervisory data; Baseline date: June 2016; Bank consolidated level data for banks headquartered in Luxembourg and sub-consolidated level data for subsidiaries of foreign banks; Market-data - Methodology and channels: - Satellite models developed by the FSAP team; Balance sheet-based approach; Market data-based approaches - Satellite models for credit losses, pre-impairment income, credit growth; expert judgment - Models to integrate solvency-funding interactions; methodology to calculate sovereign risk; methodology to calculate losses from bonds and money market instruments (haircuts based on a modified duration approach) - Net fee income and commission income projected based on assumptions on investment funds’ business volume and redemptions - Stress test horizon: 3-years (2017–2019) - Tail shocks and scenarios: - TD exercise based on baseline and adverse scenarios covering loans and trading book - TD analysis covered domestic real estate, exposures to parent companies and investment funds, and sovereign risks - Variables: domestic macrofinancial variables, GDP for key trading partners, interest rates, exchange rates and real estate prices - In the adverse scenario, the GDP growth rate declines to -3.7, -2.6 and +1.8 percent, in 2017, 2018 and 2019 respectively - Market shocks calibrated to magnitudes close to those observed in 2008/2009 - Sensitivity analysis: - Evaluated household balance sheet sensitivity to macroeconomic shocks; direct effects of interest rate shocks; exchange rate shocks; decline in sovereign bond prices; failure of the largest to 10 largest corporate exposures - Banking Sector: Liquidity Risk (Top-down by Authorities and FSAP team jointly) - Institutional perimeter: - Institutions included: 16 banks for the LCR analysis and eight for the NSFR and cash-flow analysis - Market share: 45 to 73 percent of banking sector’s assets - Data and baseline date: Latest data: September 2016 for LCR and June 2016 for NSFR and cash flow analysis; Source: supervisory data; Scope of consolidation: perimeter of individual banks - Methodology and channels: - Basel III-LCR and NSFR type proxies; Cash-flow based liquidity stress test using maturity buckets by banks; Liquidity test in foreign currencies - Risks and buffers: - Risks: Funding liquidity (liquidity outflows) and Market liquidity (price shocks) - Buffers: Counterbalancing capacity and Central bank facilities - Tail shocks: - Run-off rates calculated following historical events, IMF expert judgment and LCR/NSFR rates - Bank run and dry up of wholesale funding markets, taking into account haircuts to liquid assets - Regulatory standards: - Basel III standards (revision as of January 2013); European Commission Delegated Act - Output presentation: - Liquidity gap by bank and aggregated; Survival period in days by bank; Number of banks that can still meet their obligations - Banking Sector: Contagion Risk - Institutional perimeter: - Institutions included: 16 banks; 12 insurance companies; 5 investment fund groups covering 75 percent of the total net assets for HY, EM bond funds, VNAVs and MMFs; 54 percent for mixed funds; 32 percent for bond funds - Market share: 73 percent of total banking system assets - Data and baseline date: Latest data: June 2016; Source: supervisory and market data; Scope of consolidation: perimeter of individual institutions or sectoral indexes - Methodology and channels: - Balance sheet and off balance sheet based financial metrics; Network interbank model by Espinosa-Vega and Solé (2010); Diebold-Yilmaz variance decomposition connectedness methodology; CIMDO/SyRIN approach (Goodhart and Segoviano, 2009) - Tail shocks: - Pure contagion: default of institutions; Spillover index and transmission - Output presentation: - Number of undercapitalized and failed institutions and their shares of assets in the system; Evolution and direction of spillovers within the network ### Investment Funds: Liquidity and Solvency Risk (Top-down by FSAP Team) - Liquidity Risk — Institutional perimeter and data: - Institutions included: 191 investment funds: 42 EM bond funds; 32 HY bond funds; 40 mixed funds (investing at least 70 percent of their assets into fixed income instruments); 50 largest bond funds; 5 CNAVs short term MMFs (3 of which are in USD, 1 in EUR, 1 in GBP); 7 VNAV short term MMFs, 15 MMFs - Market share: 75 percent of the total net assets for HY, EM bond funds, CNAVs, VNAVs, and MMFs; 54 percent for mixed funds; 32 percent for bond funds - Data and baseline date: Latest data: March 2016; Source: supervisory data - Liquidity methodology and risks: - Liquidity measures by i) cash and short-term debt securities (residual maturity less than one year) and ii) cash and high quality liquid assets - Risks: Funding liquidity (liquidity outflows) and inability to sell assets to cope with redemptions - Buffers: Liquidity buffers; Credit facilities and Liquidity Management Tools - Tail shocks: - Monthly Redemption shock equal to the 1th percentile of historical net flows observed over 2007–2016 - Redemption shock estimated from an econometric model relating funds flows to macrofinancial variables (including the ones used in the bank macroeconomic scenario) - Output presentation: - Redemption coverage ratio by investment fund and liquidity shortfall; Number of funds and share of funds that cannot meet their obligations - Solvency Risk — CNAV MMFs: - Institutional perimeter: - Institutions included: 5 CNAVs short term MMFs - Market share: 75 percent of the total net assets for short term CNAVs MMFs - Data and baseline date: Latest data: March 2016; Source: supervisory data - Methodology: - Increase in risk free rates and credit spreads that would result in deviations between shadow Net Asset Value (NAV) and Constant Net Asset Value - Risks and buffers: - Risks: Inability to maintain Constant NAV - Buffers: Liquidity buffers; High credit quality and short duration assets - Tail shocks: - Sensitivity analysis with shocks to risk free rates and credit spreads ranging from 10 to 200 basis points - Regulatory guidance: - According to the IMMFA Code of Practice, draft MMF Regulation and the Ireland FSAP analysis, escalation procedures should exist for deviation between the published price and the shadow NAV above 20 basis points. *Source: Luxembourg authorities and IMF staff calculations; Thomson Reuters Datastream; Bloomberg.* ### 6. Reporting ### 6. Reporting ### Investment Funds: Results Presentation - Output presentation: - Deviations between constant NAV and shadow NAV - Number of funds and share of funds that can still meet their obligations ### Insurance Sector: Solvency Risk — Institutional Perimeter - Domain: Bottom-up by Authorities (CAA) - Bottom-up by companies under the guidance and supervision of the Luxembourg authorities - Institutions included: - 10 life insurance companies and one re-insurer - Market share: - 71.5 percent of the Luxembourg insurance market in terms of gross life technical provisions - Data and baseline date: - Supervisory data - Baseline date: June 2016 ### Insurance Sector: Channels of Risk Propagation and Methodology - Methodology: - Analysis was based on Solvency II requirements using the infrastructure developed by the Luxembourg Commission aux Assurances and/or the EIOPA - Stress test horizon: - 2017–2019 ### Insurance Sector: Scenarios and Risks - Tail shocks / Scenario analysis: - Macro scenarios included a baseline and an adverse scenario in line with the banking sector stress test - Risks / factors assessed: - Credit risk and market risk ### Insurance Sector: Reporting of Results - Output presentation: - Three-year projections of undertakings’ available own funds - The coverage of the solvency capital requirement (SCR) - Capital adequacy ratios with and without the “long-term guarantee package” --- ### Appendix IV. Methodological Assumptions for Banks’ Balance Sheet and Profit Projections ### Core assumptions - Growth of banks’ balance sheets: - Projected to grow in line with nominal GDP; size of banking system in terms of assets remains constant relative to the size of the economy. - Note: in adverse scenarios, growth of net assets and exposures at default is usually lower than growth of total assets because provisions are higher. - Projection of risk-weighted assets (RWAs): - For nine banks under Basel II Internal Rating-Based approach: risk weights projected using Basel II formula for credit risk (capital requirement ratio depends on PD, LGD and asset correlation). - For seven banks under Basel II standardized approach: RWAs projected based on new defaulted loan projections and reduced in proportion to new provisions after application of provisioning rate assumptions. - For residential mortgage segment: real provisioning rates observed in 2016 used as working assumption for projection period. - For other loan segments: provisioning rates projected so as to obtain a system-wide average provisioning rate (including mortgage) of 50 percent. - Basel II standardized approach: past due loans risk-weighted between 100 and 150 percent for unsecured portion net of specific provisions; exposures risk-weighted net of specific provisions. - Evolution of profits: - Most non-interest profit items projected to grow in line with nominal GDP. - Net fee and commission income projections incorporate investment funds’ asset growth for banks with investment-fund-related businesses. - Operational and administrative expenses assumed to grow in line with nominal GDP growth with a floor set at zero. - Income from extraordinary items assumed not to recur in 2017–2019 in baseline and adverse scenarios. - Non-performing loans assumed not to provide any accrued income. - Banks’ funding costs and lending rates: - Funding costs assumed to evolve in line with changes in the stressed Euribor rate. - For subsidiaries of foreign banks, funding-cost projections linked to intragroup funding based on the Euribor rate projected for the home country. - Lending rates (approximated by ratio of interest income to net loans) projected via an econometric OLS panel model with fixed effects (annual data 1990–2015). Explanatory variables: annual change in the 3-month money market interest rate and annual change in the amount of total funding in percent (lagged by one period). - Projection of total funding in adverse scenario aligned with liquidity stress test parameters to integrate liquidity and solvency results. - Change in interbank rate found most significant on average effective interest rate set on loans. ### Distribution of dividends - Dividend distribution rules: - i) Banks with a total Capital Adequacy ratio below the Pillar II requirements in any year of a given scenario were not allowed to distribute dividends. - ii) Banks that pass Pillar II requirements distribute dividends at a rate of 50 percent. - Note: The Pillar II requirements cannot be displayed as this is not public information for every bank. ### Empirical results for lending-rate estimation (Table A4.1) - Estimated panel fixed-effects OLS model (dependent variable: annual change in ratio of banks’ interest income to net loans) — selected outputs: - annual change in Euribor rate: 0.5893*** (t-statistic 218.64) - annual change in total funding (in percent): -0.0014*** (t-statistic -4.93) - Constant: -0.0116** (t-statistic -2.17) - R-square: 0.93 - # of observations: 255 - Note: Obtaining a robust estimate for funding costs would require much longer time series and more observations; satellite model results should be interpreted with appropriate care. *International Monetary Fund — cr17261 - 6. Reporting* ### Appendix V. Satellite Models for Credit Risk—Technical Details ### Appendix V. Satellite Models for Credit Risk—Technical Details ### Solvency stress testing methodology - Challenge: Short time series and lack of relevant crisis experience in Luxembourg (no severe banking crisis in recent decades despite a severe recession in 2009) hindered reliable estimation of credit risk parameters from Luxembourg-only data. - Approach: Apply international benchmarks and experience from advanced countries that experienced severe banking crises (Hardy and Schmieder (2013); Ireland 2008–2013) to calibrate shocks and credit loss sensitivities under extreme severity. - Starting points and adjustments: - Used Luxembourg banks’ supervisory Through-The-Cycle (TTC) PDs excluding defaulted exposures for eight broad exposure classes as the starting point. - Applied correction factors close to 1.5 to TTC PDs to approximate Point-in-Time (PIT) PDs for expected losses’ calculations. - Smoothed TTC PDs used for RWA calculation by applying a cyclical parameter of 0.5 to the annual change in projected PiT PDs. - Special treatments: - For exposures to foreign sovereigns, institutions, corporate and commercial real estate: used Moody’s/CreditEdge Expected Default Frequencies dataset on a sample of 12 advanced countries starting in 2005. - Mortgage segment: used historic PDs from countries where available in banks’ annual reports. - Basel II standardized approach exposures: used CRR standard risk weights for loan migrations and constructed a simple satellite model using NPL data. - Own sovereign exposures (for banks using advanced approach): migrated exposures by three notches in the adverse scenario and applied corresponding long-term PDs and LGDs from Moody’s report. ### Expected losses, LGD and EAD - For IRB portfolios: expected losses = proxy PIT PDs × LGD × EAD. - For standardized approach portfolios (except sovereigns): flows of NPLs were multiplied by: - the bank’s effective provisioning ratio in the mortgage segment, or - an average provisioning ratio of 50 percent for other standardized portfolios. - LGD projections: based on house price projections and bank-by-bank loan-to-value distribution. ### Data access and sources - Supervisory data availability: - Data at the individual bank level: fully available (in a physical data room only). - Data aggregated along groups of banks: partially available. - Data aggregated at the banking system level: fully available. - Other data: publicly-available data, commercial providers, EBA Transparency exercise, banks’ annual reports. - Note: supervisory data for estimation and testing was available at individual bank level, on solo, consolidated and sub-consolidated bases in a physical data room only. ### PD projection modelling (panel-based approach) - Dependent variable: logit transform of PDs (to ensure predictions between 0 and 1 and capture nonlinearities). - Logit transformation applied to original PD: Y_it = ln(PD_it / (1 - PD_it)). - Model specification (typical form): - Y_i,t = α + β X_t + δ Y_i,t-1 + ε_i,t, where Y_i,t is logit(PD) for asset class i at time t, X_t vector of macrofinancial regressors, and Y_i,t-1 lagged dependent variable. - Determinants by exposure class (as used in regressions): - Retail household mortgage loans: contemporaneous year-on-year Luxembourg real GDP growth (in log); real lending rate on loans to borrowers (lending rate) lagged by one quarter. - Other retail (consumer) loans: year-on-year euro area real GDP growth (in log) lagged by one period; lending rate on consumer loans lagged by one quarter; include lagged dependent variable to correct for serial autocorrelation. - Retail SME loans secured by immovable property: year-on-year euro area real GDP growth (in log); real lending rate on corporate loans; year-on-year change in commercial real estate prices; euro-dollar exchange rate. - Unsecured retail SME loans: year-on-year euro area real GDP growth (in log) lagged by two quarters; real lending rate on corporate loans; dependent variable lagged by one quarter. - Corporate SME loans: year-on-year euro area real GDP growth (in log) lagged by one period; lending rate on corporate loans lagged by one quarter; dependent variable lagged by one quarter. - Non-financial corporate loans: year-on-year euro area real GDP growth (in log) lagged by one period; real lending rate on corporate loans lagged by six quarters; euro-dollar exchange rate. - Specialized lending corporate loans: year-on-year change in the lending rate on corporate loans lagged by three quarters; dependent variable lagged by one quarter. - Loans to financial institutions: year-on-year euro area real GDP growth (in log) lagged by one period; real Euribor rate lagged by two quarters; euro-dollar exchange rate. - Sovereign exposures: dependent variable = year-on-year change in Moody’s/CreditEdge Expected Default Frequency; regressors = year-on-year change in sovereign rate lagged by one period and lagged dependent variable (to avoid endogeneity). - Basel II STA exposures: dependent variable = logit transform of non-performing loans ratio; regressors = contemporaneous Luxembourg real GDP growth and lagged dependent variable. - Estimation sample and period: projections of PIT PDs are based on quarterly data over the period 2005Q1-2016Q2, applying international benchmarks from advanced countries that experienced banking crises (e.g., Ireland). ### Use of alternative satellite models and benchmarking - ECB approach: Bayesian Model Averaging (BMA) technique at country and portfolio level; dependent variable = default rate at country and segment level; produces PD multiples attached to starting PIT PDs. - IMF Top-Down model: used historic PDs and LGDs from banks’ annual/Pillar 3 reports where available; used Moody’s/CreditEdge EDF data otherwise (commercial real estate, corporate, financial institutions, sovereign classes). - PD/NPLs and LGD projections from IMF compared with ECB benchmarks to minimize model error risks. ### Key estimation results and sensitivities (selected) - Sensitivities reported for major segments: - Corporate SME loans: a 1 percentage point increase in the lending rate → PD rises by 4.7 percentage points; a 1 percentage point decrease in euro area GDP growth → PD rises by 2.3 percentage points. - Household mortgage loans: a 1 percentage point rise in the real lending rate → PD increases by 4.6 percentage points; a 1 percentage point decline in Luxembourg GDP growth → PD increases by 2.1 percentage points. - Non-financial corporate loans: a 1 percentage point reduction in euro area GDP growth → PD increases by 2 percentage points; a 1 percentage point increase in the real lending rate → PD increases by 0.4 percentage point. - NPL model: a 1 percentage point decline in Luxembourg GDP growth → NPL ratio increases by 3.7 percentage points. - Table A5.2 (regression summary; dependent variable = logit(PD) except sovereigns): - Dependent variable lagged by 1 period coefficients (selected): HH Mortgages = -0.5032*** (t = 9.62); Other retail = 0.7414*** (t = 6.65); Corporate SMEs = -0.6049*** (t = 8.12); Institutions = 0.2068*** (t = 7.02). - YoY real GDP growth (in log) coefficients (selected): HH Mortgages = -8.4138*** (t = -6.9); Institutions = -8.0412*** (t = -8.42); Sovereigns = -24.3615*** (t = -7.63). - Real lending rate (in percent) (lagged) coefficients (selected): HH Mortgages = 0.1972*** (t = 12.1); Other retail = 0.1509 (t = 1.15). - Real Euribor rate (in pp) (lagged by 2 periods): Institutions = 0.1843*** (t = 3.52). - EUR/USD exchange rate level: Corporate = 1.6414*** (t = 4.43); Institutions = -2.1758* (t = -1.94). - Constants and model fit: R-square ranges reported (e.g., HH Mortgages = 0.75; Other retail = 0.91; Corporate SMEs = 0.96; NPLs = 0.32). - Number of observations by equation: 41, 79, 83, 54, 54, 44, 35 (as tabulated). - Significance notation: * at the 10 percent level; ** at the 5 percent level; *** at the 1 percent level. - Interpretation: Luxembourg corporate portfolio segment is most sensitive to macrofinancial variables; lending rate changes have notably large effects on PDs for corporate SME and household mortgage segments. --- ### Appendix VI. Household Stress Tests—Technical Details ### Data and updating - Source: 2014 Eurosystem Household Consumption and Finance Survey (HCFS) sample for Luxembourg — detailed data on about 1,600 households (income, financial and non-financial assets, monthly payments for total debt and primary residential mortgage). - Data collection: 2014 (prior to FSAP stress-test start). To approximate household financial situation in 2016 before stress, income and expenditure data were updated with aggregate national accounts figures. ### Stress-period projections (2017–2019) - Two econometric models estimated to project annual wage and household consumption growth using real GDP growth and inflation as explanatory variables. - Scenario parameters (Table 6.1: Dynamics of household stress test parameters, 2016–2019, in percent): - Growth in expenditure: 2016 = 2.5% ; 2017 = 0.9% ; 2018 = 1.1% ; 2019 = 2.1% - Growth in social transfers: 2016 = 0% ; 2017 = -5% ; 2018 = -5% ; 2019 = -5% - Growth in wages: 2016 = 3.7% ; 2017 = -5% ; 2018 = -5% ; 2019 = -5% - Unemployment rate: 2016 = 6.4% ; 2017 = 9.1% ; 2018 = 11% ; 2019 = 10.5% - Interest rate shock (change in lending rate for house purchase, in ppts): 2016 = 0 ; 2017 = 1.3 ppt ; 2018 = 0.6 ppt ; 2019 = 0.4 ppt - Growth in house prices: 2016 = 5.4% ; 2017 = -17% ; 2018 = -15% ; 2019 = -1% ### Household sample descriptive statistics (pre-stress) - Share of households with debt and receiving social transfers: 23% - Share of households unemployed (no member employed): 6.3% - Share of households with negative financial margin: 9% - Pre-shock defaulted households in sample: around 2.5% (higher than average PDs of banks’ mortgage portfolios). ### Financial margin, net assets and implied PD calculation - Financial margin (FM) computed from annualized income and expenditure data: - All income and expenditure data annualized. - Regular social transfers excluded pensions. - Current expenditure excluded residential mortgage and other debt payments (to avoid double counting). - Treatment of households with negative FM: - Such households assumed to first use other resources (savings), borrow from family/relatives, and finally sell real estate before default. - Net assets (NA) computed with shocked growth rates for real estate prices (μ) and financial assets (θ). If FM < 0 and NA < 0 (liabilities exceed value of real and financial assets), household assumed in default position. - Implied PD formula used (as presented in the source): PD_implied = ... (pre-shock implied PD reported as around 2.5%). - Caveat: simplification — households might access other resources (family support) even with negative NA. ### Household PDs and loan distribution findings - Distribution features: - Debt incidence by income decile has an inverted U-shape: more households in higher income deciles have debt than in lower deciles up to the 4th decile; beyond the 4th decile proportion with loans declines. - PDs are highest in the 1st, 4th and 3rd deciles; highest income deciles have highest proportion of debt and lowest PDs. In the sample, decile 9 reported no defaults. - Figures referenced (A6.1 and A6.2) show: - Distribution of household loans and implied PDs by income decile, pre-stress period. - Distribution of household PDs by income decile, before and after shocks. - Stress impact (qualitative): Under the adverse scenario, the average implied PD for household loans increases (explicit post-shock PD values and distributions are shown in the referenced figures). *Source: IMF staff calculations (Appendix V and Appendix VI text and tables as provided).* ### 2.5 to 6.4 percent. These projections are in line with those obtained from macroeconometric satellite ### cr17261 - 2.5 to 6.4 percent. These projections are in line with those obtained from macroeconometric satellite ### Mortgage debt concentration and projection note - The share of outstanding balance of mortgage debt for the 1st decile (households with highest PD) is equal to 4.1 percent of total outstanding balance of mortgage debt. - Projections discussed: "2.5 to 6.4 percent. These projections are in line with those obtained from macroeconometric satellite models used in the banking sector stress test using Irish benchmarks and with the Central Banks of Luxembourg’s own estimates." ### Appendix VII — Solvency Stress Test for CNAV MMFs: Methodology - Objective: estimate impact of increase in risk-free rates and credit spread on mark-to-market value of CNAV (‘shadow’ NAV). - Portfolio duration: for each CNAV, duration of the portfolio is computed using data on each security holding. - Change in NAV formula (as presented): change in NAV for a change in yield is calculated as the product of the shock times the duration. - Treatment of holdings: - Sovereign holdings: subject to increase in the risk-free rate. - Other securities: subject to both an increase in the risk-free rate and in credit spreads. ### Appendix VIII — Liquidity Stress Test for UCITS Funds: Calibration of the Redemption Shock (Historical Approach) - Data: monthly net flows over January 2007–March 2016 for each fund in the sample. - Redemption shock (historical approach): calculated at the 1 percent highest outflows (in percent of TNA) observed during the sample period. - Note: while monthly data used, worst weekly outflows typically range between 10 and 20 percent. ### Table A8.1 / Historical summary statistics (selected fund strategies) - EM: Average 18%, 1st Quartile 10%, Median 16%, 3rd Quartile 26% - HY: Average 19%, 1st Quartile 13%, Median 17%, 3rd Quartile 23% - Mixed funds: Average 9%, 1st Quartile 3%, Median 7%, 3rd Quartile 13% - Other bond funds: Average 18%, 1st Quartile 8%, Median 17%, 3rd Quartile 21% - Short Term CNAV: Average 19%, 1st Quartile 15%, Median 17%, 3rd Quartile 23% - Short Term VNAV: Average 23%, 1st Quartile 19%, Median 23%, 3rd Quartile 27% - Other MMFs: Average 18%, 1st Quartile 13%, Median 15%, 3rd Quartile 22% ### Calibration of the Redemption Shock (Forward-Looking Approach) - Econometric model: monthly net flows at fund-strategy level (EM, HY, Mixed and other bond funds) related to macrofinancial variables from macroeconomic model used in banking sector stress test. - Macrofinancial explanatory variables: - 3-Month Euribor - euro area 10-year rate - Eurostoxx 50 - VIX index - Term spread (euro area 10-year rate minus 3-Month Euribor) - HY bond spread (used for HY funds net flows and mixed funds) — Bank of America Merril Lynch index - EM bond spread (for EM and other bond funds) — Bank of America Merril Lynch index - Variable treatment: included in differences (3-Month Euribor, 10-year rate and term spread) or variations (Eurostoxx50, VIX index, bond spreads). - Model performance: Figure 34 indicates models replicate past evolutions for fund types; MMFs excluded from forward-looking approach because explanatory variables were not significant. - Forecasting under adverse scenario: - Baseline scenario values as of 2016Q4 used; projected values derived from 2017Q1 adverse scenario values converted to monthly frequency assuming constant monthly change. - HY and EM spreads not included in adverse scenario; assumed subject to highest monthly increase observed since 2007. - Table A8.2 (presented values for explanatory variables under adverse scenario): - 3M-Euribor: Initial -0.33, Forecast -0.13, Change 0.20 - EA 10-Year: Initial 1.00, Forecast 1.32, Change 0.33 - Term spread: Initial 1.33, Forecast 1.45, Change 0.13 - Eurostoxx 50: Initial 100, Forecast 94, Change -6% - VIX: Initial 14.80, Forecast 23.44, Change 58% - EM spread (bps): Initial 47, Forecast 88, Change 380 - HY spread (bps): Initial 49, Forecast 107, Change 6578 - Projected net flows under adverse scenario (Table A8.3 summary, in percent of TNA): - EM: Net flows 9% - HY: Net flows 11% - Mixed funds: Net flows -11% (negative indicates net inflows under adverse scenario) - Other bond funds: Net flows 6% ### Measure of Liquid Assets: Short-Term Assets - Metric: cash and debt securities with residual maturity of less than one year (similar to ESRB risk dashboard measure). - Data source: BCL regular collection. - Limitation: excludes securities with maturity > one year, even investment grade sovereign bonds; alternative measure applied based on HQLA approach used for banks (LCR). ### Measure of Liquid Assets: High-Quality Liquid Assets (HQLA) approach - Sample: 191 funds; full portfolio of securities obtained; 21,911 securities analyzed; 2,667 securities excluded due to data gaps. - Final sample coverage: covers 88 percent of the securities and 84 percent of the portfolio of securities. - Coverage note: All funds except two had a coverage of at least 95 percent of their TNA. For funds with missing data, adjusted-HQLA robustness checks were run and stress test results were similar. - Liquidity weights assigned following ESMA (2015) and Basel Committee LCR weights; cash weight = 100 percent; non-investment grade corporate bonds weight = 0 percent. - Liquidity weight table (selected entries by instrument and rating): - Sovereign bonds AAA to AA-: 100% - Corporate bonds AAA to AA-: 85% - Securitization AAA to AA-: 85% - Equities AAA to AA-: 100% (as presented) - Sovereign bonds A+ to A-: 85% - Corporate bonds A+ to A-: 50% - Securitization A+ to A-: 50% - Equities A+ to A-: 85% - Sovereign bonds BBB+ to BBB-: 50% - Corporate bonds BBB+ to BBB-: 50% - Securitization BBB+ to BBB-: 0% - Equities BBB+ to BBB-: 50% - Below BBB-: Sovereign 0%, Corporate 0%, Securitization 0% - Cash and short-term debt HQLA weight: 100% (implied by cash weight) - Aggregated HQLA by fund type (Table A8.5 averages, in percent of TNA): - Short Term CNAV: 98% ^ - Short Term VNAV: 99% ^ - Other MMFs: 88% ^ - EM: 16% (short-term assets) and 42% (HQLA) — layout as presented - HY: 12% (short-term assets) and 12% (HQLA) - Mixed funds: 31% (short-term assets) and 55% (HQLA) - Other bond funds: 20% (short-term assets) and 56% (HQLA) - ^ symbol retained as in source presentation ### Assessment of Resilience of Individual Funds - Redemption Coverage Ratio (RCR) formula (as presented): - RCR = Liquid assets / Net outflows - Liquidity shortfall for funds with RCR below one (formula as presented): - Liquidity shortfall = Net outflows − Liquid assets - All quantities in percent of TNA. ### Deposit Outflows from Funds: Waterfall and Prorata liquidation approaches - Waterfall approach: fund uses most liquid assets (HQLA securities) first, then cash if HQLA insufficient. - Prorata approach: liquidates HQLA securities and cash proportional to composition of liquidity buffers; if proportional liquidation insufficient, remaining cash is used. - Results presented are averages of waterfall and prorata approaches. ### Table A8.6 — Deposit Outflows (in EUR billion and percent of TNA) - Total deposits: - Including mixed funds: 24.7 - Without mixed funds: 21.6 - Deposit outflows (waterfall): - Including mixed funds: 3.0 - Without mixed funds: 3.3 - Deposit outflows (waterfall, in % of fund deposits): - Including mixed funds: 12% - Without mixed funds: 15% - Deposit outflows (prorata): - Including mixed funds: 6.8 - Without mixed funds: 7.2 - Deposit outflows (prorata, in % of fund deposits): - Including mixed funds: 28% - Without mixed funds: 33% - Deposit outflows (average of waterfall and prorata): - Including mixed funds: 4.9 - Without mixed funds: 5.3 - Deposit outflows (average of waterfall and prorata in % of funds deposits): - Including mixed funds: 20% - Without mixed funds: 24% *International Monetary Fund. Content derived from the referenced chapter and appendices in the provided PDF.* --- _Source: https://www.imf.org/-/media/files/publications/cr/2017/cr17261.pdf_