## 1belea2023011 ## Source details **Canonical URL:** [1belea2023011](https://www.imf.org/-/media/files/publications/cr/2023/english/1belea2023011.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2023/english/1belea2023011.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2023/english/1belea2023011.pdf.json) --- ### Executive summary — overarching assessment - Financial sector resilient to a series of shocks; well capitalized and profitable, but risks remain. - Bank profitability and capital have surpassed pre-pandemic levels; liquidity buffers remain strong; banks’ direct exposures to Russia limited. - Insurance and investment funds sectors weathered the pandemic well. - Limited evidence of credit or asset price booms over the past decade. - Inflationary pressures (partly from spillovers of Russia’s war on Ukraine) and rapid tightening of financial conditions have weakened activity and reined in credit demand. - GDP growth projected to slow in 2023, before returning to potential over the medium-term. ### Macroeconomic outlook and scenario paths - Adverse scenario: intensifying fallout from the war in Ukraine, worsening energy crisis and supply disruptions, higher inflation and monetary tightening, housing market corrections and broader macrofinancial channels. - Adverse scenario GDP growth (Real GDP growth, percentage): -3.8 (2023), -2.2 (2024), 1.1 (2025), 4.4 (2026). - Selected macro paths (Stress): - Unemployment (percentage): 6.4 (2023), 9.1 (2024), 9.3 (2025), 7.1 (2026). - Inflation (percentage): 11.4 (2023), 3.9 (2024), 1.1 (2025), 2.6 (2026). - 3 months EURIBOR (percentage, year average): 6.2 (2023), 4.9 (2024), 3.2 (2025), 2.7 (2026). - House Price Index (2022=100): 75.6 (2023), 71.0 (2024), 75.8 (2025), 81.7 (2026). - CRE Price Index (2022=100): 70.9 (2023), 66.3 (2024), 78.2 (2025), 85.6 (2026). - Equity prices (Index, 2022=100): 71.2 (2023), 66.4 (2024), 74.1 (2025), 83.2 (2026). - 10-year Belgian Sovereign Bond Yield (percentage): 6.1 (2023), 5.0 (2024), 2.9 (2025), 2.5 (2026). ### Banking sector — solvency stress test (aggregate findings) - Sample: seven Significant Institutions (SIs) covering over 90 percent of banking sector assets; top-down FSAP exercise using Q4 2022 supervisory data. - Starting aggregate CET1 capital ratio: 18.3 percent (end-2022). - Adverse scenario outcome: - Aggregate CET1 ratio drops to 14 percent (end-2026) from 18.3 percent. - All banks meet minimum capital requirements; one bank fails to meet its CcoB/O‑SIIB in the adverse scenario. - Four-year cumulative credit impairments under stress: 108.3 percent of starting CET1 capital by end-2026. - Baseline four-year cumulative impairments: 48.9 percent of starting capital. - Profitability and NII: - ROA falls from 58bps in year 0 to 28bps in year 1 under stress; losses recorded in remaining three years. - Average NII over total assets in the adverse scenario rises to 2.4 percent by end-2023 (scenario 3-months EURIBOR increases to 6.2 percent), then average NII ratio drops to 1.5 percent at end-2025. - Market risk losses: four-year cumulative market losses 0.9 percent of starting CET1 under stress (0.2 percent in baseline). - Heterogeneity: variation in credit provisions/impairments linked to average LTV ratios and lending portfolio composition. ### Sensitivity analysis — deposit conversion (ALM) - Design: convert non-term deposits to term deposits to simulate depositor response to rising interest rates; credit and market losses held constant to isolate ALM effects. - "20 percent transition" results (four-year horizon): - CET1 ratio trajectory from 18.3 percent to 13.1 percent (final year). - NII over total assets: 2.3 percent (year one) to 1.4 percent (year four). - ROA: 0.3 percent (year one) to -0.6 percent (year four). - One bank fails to meet minimum CET1 requirements by the fourth year. - "50 percent transition" results (four-year horizon): - CET1 ratio trajectory from 18.3 percent to 11.5 percent (final year). - NII over total assets: 2.0 percent (year one) to 1.4 percent (year four). - ROA: negative from second year onwards, ranging to -0.6 percent by fourth year. - The same bank as in the 20 percent scenario fails to meet minimum CET1 requirements by the fourth year. - Comparative finding: depositor behavior materially affects ALM, profitability and capital adequacy; supervisory stress tests should incorporate depositor behavior. ### Liquidity stress tests (LCR, cash flow, NSFR) - LCR stress testing: - Starting LCR around 150 percent (system level). - System remains generally liquid in most scenarios except the most severe; two out of seven banks fall below regulatory hurdle of 100 percent in most scenarios. - When Basel coefficients are doubled, system-wide LCR falls below the regulatory threshold. - Cash-flow analysis (12-month horizon): - Initial counterbalancing capacity approximately 24 percent of total bank assets; diminishes by factor of four after 12 months but system remains liquid overall. - Deposit runoff assumptions (cumulative over three months, percent): - Stable retail deposits: 5-10 - Other retail deposits: 20-25 - Operational deposits: 25-40 - Non-operational deposits from fin. customers: 30-50 - Non-operational deposits from non-fin. & other customers: 30-50 - Committed credit facilities: 60-12 - Liquidity facilities: 75-75 - Haircut ranges on counterbalancing capacity: - Low stress: between 7 and 30 percent - High stress: between 15 and 40 percent - Two out of seven banks encounter liquidity gaps mostly due to deposit outflows; frontloaded outflows reduce survival time to three days for affected banks. - USD liquidity weaker for banks with USD exposure: four of seven banks have USD exposure; two of these show USD liquidity shortages. - NSFR (qualitative): - NSFR at end-2022: 132 percent (qualitative test highlights heavy reliance on stable funding, mostly deposits covered by public guarantee). - Debt securities (excluding CDs) about 6 percent of total liabilities; deposits from households and NFCs >50 percent of liabilities. ### Solvency–liquidity interaction (forced HTM liquidation) - HTM portfolio: 7.8 percent of bank assets (approx. 73 percent of total debt securities held). - HTM composition: sovereign debt 65.5 percent; debt from credit institutions 20.6 percent; corporate debt 7.4 percent; non-bank financial institutions 6.1 percent; central banks 0.4 percent. - Four forced-liquidation scenarios (aggregate fractions sold): Market 47.4 percent; Outflow 66.6 percent; Combined 86.6 percent; Complete liquidation 100 percent. - Year 1 pre-liquidation: average CET1 18.3 percent; total portfolio devaluation 7 percent. - Losses as proportion of CET1 before provisions reversal: Market 453bps; Outflow 636bps; Combined 826bps; Complete liquidation 955bps. - Reversal of provisions: 31bps to 95bps. - Net losses over pre-liquidation CET1 after provisions reversal: Market 421bps; Outflow 626bps; Combined 766bps; Complete liquidation 860bps. - RWAs reduction: 2 percent. - Post-liquidation CET1 ratios (after 2 percent RWA reduction): Market 18.4 percent; Outflow 18.3 percent; Combined 18.2 percent; Complete liquidation 18.1 percent. - Outcome: all banks remained above minimum capital requirements. - Year 3 pre-liquidation: average CET1 15.4 percent; total portfolio devaluation 4.8 percent. - Losses as proportion of CET1 before provisions reversal: Market 364bps; Outflow 511bps; Combined 663bps; Complete liquidation 766bps. - Reversal of provisions: 37bps to 111bps. - Net losses after provisions reversal: Market 328bps; Outflow 500bps; Combined 592bps; Complete liquidation 655bps. - Post-liquidation CET1 ratios (after 2 percent RWA reduction): Market 15.3 percent; Outflow 15.2 percent; Combined 15.1 percent; Complete liquidation 15.0 percent. - Post-liquidation ROA: -78bps, -82bps, -83bps, and -85bps respectively. - Outcome: all banks continued to meet minimum capital requirements. ### Insurance sector — solvency and liquidity stress testing - Sample and scope: - Eight composite insurers: KBC, Belins, Ethias, Axa, Baloise, AG Insurance, Allianz, and P&V — covering around 76 percent of Belgian insurance market. - Aggregated balance sheet assets of participating insurers: €231 billion. - Solvency stress test outcomes (median): - Pre-stress median solvency ratio: 192 percent. - Post-stress median solvency ratio (without management actions): 113 percent. - Post-stress median solvency ratio (including reactive management actions): 142 percent. - Aggregated solvency surplus: €9.1 billion before stress; €1.8 billion after stress; €3.3 billion with reactive management actions. - Role of Volatility Adjustment (VA): - VA substantially mitigates spread shock for insurers applying VA; without VA most insurers’ solvency ratios would have fallen below 100 percent. - Recommendation to run a detailed assessment of the impact of the use of the VA on solvency ratios in a changing interest rate environment. - Key vulnerability channels: - Corporate bond exposures: 15 percent of total assets; significant share in BBB grade — a rating migration akin to 2008 would reduce solvency ratio by about 14 percentage points. - Belgian sovereign exposure: 18 percent of balance sheet; a further 30 basis points shock reduces solvency ratio by nine percentage points (~€1 billion valuation loss). - Real estate exposure: about ten percent of insurers’ balance sheets; stressing all real estate exposures reduces EOF by 19 percent on average. - Tiering and EPIFP concerns: unrestricted tier 1 ~80 percent of eligible own funds; EPIFP represents significant share of unrestricted tier 1 for several insurers, raising loss-absorbency concerns. - Liquidity stress (variation margin and lapses): - IRS exposures concentrated in few insurers; weighted average remaining time to maturity: 12.6 years (floating-for-fixed) and 13.8 years (fixed-for-floating); open positions: 181 and 519 respectively. - Variation margin stress: modeled shocks of +25bps, +50bps, +100bps over short horizons; for a 25-bps overnight move cash margin calls could be met by drawing on 22 percent of cash equivalents; for >100bps would require 82 percent. - Mass lapse scenarios: - 30 percent mass lapse for all lapsable contracts; variations targeted at high-risk contracts; combination scenario: 10 percent withdrawal for products with penalties plus 30 percent for high-risk products. - High-risk contracts average share: 14 percent of lapsable surrender values (range 4–39 percent). - Liquidity finding: insurers generally have sufficient liquid funds to withstand significant redemptions, but dispersion is high; some insurers may need to liquidate high-quality government bonds or access repos, which may be unavailable in crisis. - Management actions: - Reactive management actions reduce solvency impact by about 20 percentage points on median solvency ratio; typical actions include dividend retentions, hedging, asset sales, reinsurance, intra-group capital moves. ### Investment funds — liquidity stress testing and market impact - Sample: 435 funds (end-2022) with NAV around €136 billion (sample includes Bond 8%, Mixed (FoF) 58%, Non-public AIF 12%, Other 22%). - Assets under management (sample used in stress test susceptible to run): approx. 153 EUR bn at end-2022 (in exercise subset). - Objectives: assess resilience to severe but plausible redemption shocks; identify vulnerable fund types; estimate capacity to transmit shocks. - Methodology: - Redemption shock calibration via historical VaR and ES (focus on ES 3% main shock), homogeneous and heterogeneous calibrations. - Liquid asset measurement using Basel III HQLA weights; Redemption Coverage Ratio (RCR) = Highly Liquid Assets / Redemption shock. - Liquidation strategies: vertical slicing (pro rata) and waterfall (sell most liquid first). - Price impact based on market depth using average daily trading volumes and volatility; τ = 1 day for instantaneous shock. - Key results: - Sector largely able to withstand severe but plausible redemption shocks. - Less than 2 percent of funds (6.7 percent of NAV) analysed would not have enough HQLA to meet redemptions in market stress and thus present liquidity shortfalls (Heterogeneity ES 3%: Total 7 funds; 1.6% of funds; 6.7% of NAV with RCR < 1). - Non-public alternative funds particularly vulnerable: under Homogeneity ES 3% five funds with RCR < 1 (% Funds 13%; % NAV 46.3%). - FoFs hold >40 percent of total NAV and are largely invested in other CIUs including ETFs domiciled in other European jurisdictions. - Unencumbered cash: 2 percent of total AuM. - Fixed income exposure: €34 billion equally split between sovereign and corporate bonds; >80 percent corporate bonds are investment grade. - Market impact and contagion: - Liquidation of domestic sovereign bonds by funds would amount to 10-13 percent of assets liquidated by investment funds in reference scenarios; market depth sufficient to avoid relevant price impact under tested liquidation strategies. - Volatility spillovers: Belgian funds more vulnerable to volatility shocks originating in other European jurisdictions and global markets than to domestic market shocks; mixed vehicles show stronger co-movements and inward spillovers from foreign markets. ### Interconnectedness and contagion analysis - Domestic interbank contagion (Espinoza and Solé (2010) framework, large exposures COREP data end‑2022): - Parameters: Lambda = 65 percent (credit shock), Delta = 50 percent (funding shock), Rho = 35 percent. - Result: failure of a single domestic bank would not trigger failure of another bank; contagion vulnerability low overall though a few banks are systemically important in interbank network. - Cross-border contagion: - Belgian banks’ foreign claims equal 50 percent of GDP; exposures concentrated towards Czech Republic, UK, Netherlands, France, Slovakia (>60 percent of cross-border). - Cross-border parameters: Lambda = 65 percent; Delta = 50 percent; Rho = 45 percent. - Simulation: Belgium highly vulnerable to contagion from external shocks — defaults in Czech Republic, Netherlands, France, Luxembourg, and UK would have significant adverse impact on Belgian banks’ capital. - Cross-sector market interconnectedness: - Iterative regression-based mapping of solvency shocks into yield and share-price changes converges after about 12 iterations. - Losses from crossholdings limited: banks’ equity-to-assets ratio declines from 4.267 percent to 4.266 percent; insurers’ equity-to-assets ratio declines from 5.913 percent to 5.777 percent. - Insurers’ larger decline explained by exposures to revaluation losses from holdings of bank-issued debt (~€1 billion exposures to banks’ debt; ~€17 billion to investment funds’ issued shares). ### Key recommendations (high level) - NBB: strengthen General Stress Testing Framework by integrating individual models, incorporate IFRS 9 into Credit Risk Modeling, and adopt advanced analytical methods for ALM risks (Timing: MT). - NBB: increase insurers’ resilience to macrofinancial shocks by engaging industry to reduce dependence on lower tier capital and implement liquidity stress tests and scenario analysis (Timing: MT, NT). - NBB: run detailed assessment of the impact of the volatility adjustment on insurers’ solvency ratio in a changing interest rate environment (Timing: NT). - FSMA & NBB: establish formal data-sharing agreement on investment funds’ portfolio holdings and supervisory data; FSMA to develop/adapt a stress test framework for asset management and NBFI and monitor non-public open-ended AIFs and Liquidity Management Tools (LMTs) (Addressees: FSMA; NBB; Timing: C, MT). *Source: EXECUTIVE SUMMARY and selected sections, 1belea2023011.* ### EXECUTIVE SUMMARY __________________________________________________________________________ 8 ### 1belea2023011 - EXECUTIVE SUMMARY ### Overall assessment and macrofinancial context - The financial sector has remained resilient to a series of shocks and is well capitalized and profitable, but risks remain. - Bank profitability and capital have surpassed pre-pandemic levels. - Liquidity buffers remain strong and banks’ direct exposures to Russia are limited. - The insurance and investment funds sectors have also weathered the pandemic well. - There is limited evidence of credit or asset price booms over the past decade. - Inflationary pressures, partly the result of spillovers from Russia’s war on Ukraine, and a rapid tightening of financial conditions have weakened activity and reined in credit demand. - Among high uncertainty, GDP growth is projected to slow in 2023, before returning to potential over the medium-term. ### Banks: solvency stress test (aggregate findings) - Adverse scenario assumptions: intensifying fallout from the war in Ukraine, worsening energy crisis and supply disruptions, and higher inflation leading to monetary tightening; housing market corrections and broader macrofinancial channels affecting asset valuations, funding costs, interest rate risks, and credit risk. - The adverse scenario commences with two consecutive years of recession, with real GDP growth rates of -3.8 percent and -2.2 percent respectively. - Under this severe but plausible scenario: - The aggregate common equity tier 1 (CET1) ratio of the group of stressed banks drops to 14 percent from 18.3 percent, still well above regulatory requirements. - While some heterogeneity between banks exists (mainly in ALM and NII projections), all banks can meet minimum capital requirements. ### Banks: sensitivity analysis on deposit conversion - A sensitivity analysis converted non-term deposits to term deposits owing to the steep surge in interest rates. - Scenario results over four years: - 20 percent conversion revealed a CET1 ratio trajectory from 18.3 percent to 13.1 percent. - 50 percent conversion revealed a CET1 ratio trajectory from 18.3 percent to 11.5 percent. ### Liquidity stress tests (LCR, cash flow, NSFR) - Liquidity risk was evaluated through three types of analyses: Liquidity Coverage Ratio (LCR) stress tests, cash flow-based analysis, and descriptive analysis of the Net Stable Funding Ratio (NSFR). - LCR analysis shows Belgian banks are more susceptible to stress originating from the retail and wholesale sectors, while they are relatively insulated from market-driven stress. - Cash-flow analysis shows banks are well protected against market-induced stress, but some banks are vulnerable to liquidity strains caused by deposit outflows. - The qualitative NSFR test reveals heavy reliance on banks’ stable funding sources, consisting mostly of deposits covered by the public guarantee. - Overall: liquidity levels are comfortable for the system but need to be reinforced for some banks. ### Solvency–liquidity interaction (forced liquidation analysis) - Framework: simulates severe liquidity stress forcing divestment of Held-To-Maturity (HTM) assets at market value, causing devaluation. - Analysis encompassed four scenarios determined by the fraction of the HTM portfolio sold. - Under all four scenarios CET1 ratios remained between 15.3 percent and 15 percent by the end of year 3 (2025), when banks record their highest losses. - All banks still meet their minimum capital requirements. ### Insurance sector stress test (solvency and liquidity) - Solvency stress test: insurance industry generally able to withstand the severe scenario, although solvency ratios fall significantly if management actions are ignored. - Median solvency ratio outcomes: - Pre-stress median solvency ratio: 192 percent. - Post-stress median solvency ratio (without management actions): 113 percent. - Post-stress median solvency ratio (including reactive management actions): 142 percent. - The analysis showed scope to gradually improve the quality of insurers’ capital. - Liquidity stress analysis: - Low derivative exposures for most insurers make the sector largely resilient to variation margin calls following steep interest rate hikes. - Insurers generally have sufficient liquid funds to withstand significant redemptions, but results show high levels of dispersion. ### Investment funds liquidity stress test - Objective: assess resilience of the Belgian fund industry to severe but plausible redemption shocks. - Findings: - The sector largely would be able to withstand severe but plausible redemption shocks. - Less than 2 percent of the investment funds analysed would not have enough highly liquid assets to meet investors’ redemption requests in a market stress situation and thus present liquidity shortfalls. ### Interconnectedness analysis - Domestic and cross-border interconnections are relatively modest. - Domestic interbank market: low levels of contagion risks for Belgian banks, though a few banks exhibit a high degree of systemic importance within the interbank system. - Cross-border analysis: reveals Belgian banks’ strong exposures to non-financial sectors. ### Key recommendations (Table 1: FSAP Systemic Risk Analysis Key Recommendations) - 1. Further strengthen the National Bank of Belgium’s (NBB) General Stress Testing Framework by (i) integrating individual models within the stress testing framework to offer a comprehensive impact analysis on banks' profitability and capital adequacy, (ii) incorporating the IFRS 9 Approach into Credit Risk Modeling and (iii) continuing to adopt advanced analytical methods to better monitor Assets Liabilities Management risks. - Addressee: NBB - Timing: MT - 2. Increase insurers’ resilience against macrofinancial shocks, by i) engaging with industry to reduce dependence of insurers on lower tier capital and ii) implementing liquidity stress tests and scenario analysis to identify potential sources of stress. - Addressee: NBB - Timing: MT, NT - 3. The National Bank of Belgium (NBB) to run a detailed assessment of the impact of the use of the volatility adjustment on the insurance firms’ solvency ratio in a changing interest rate environment. - Addressee: NBB - Timing: NT - 4. The NBB and the Financial Services and Markets Authority (FSMA) to establish a formal agreement for sharing data on investment funds’ portfolio holdings and supervisory data and continue common reporting on asset management and Non-bank Financial Institutions (NBFI). The FSMA to (i) develop and adapt a stress test framework to assess the structural vulnerabilities stemming from the investment fund sector and integrate into the NBB’s systemic risk assessment, (ii) improve their data driven risk-based supervisory framework by using this enhanced information set. (iii) monitor more closely Belgian managers of non-public open-ended AIFs, to enhance their liquidity management including by recommending the availability and use of Liquidity Management Tools (LMT) where necessary, (iv) monitor and assess the use of LMTs by investment funds (IFs) within the European framework. - Addressee: FSMA; NBB - Timing: C, MT *Source: EXECUTIVE SUMMARY, 1belea2023011 - EXECUTIVE SUMMARY.* ### 1. Economic activity has weakened, inflation remains high, and the fiscal outlook is ### 1. Economic activity has weakened, inflation remains high, and the fiscal outlook is challenging ### Recent economic developments and outlook - Indirect spillovers from Russia’s war on Ukraine have softened demand and growth has weakened. - Automatic wage indexation and government measures have supported households. - Uncertainty remains elevated and rapidly tightening financial conditions are likely to dampen credit and activity going forward. - GDP growth is projected to slow in 2023, before returning to potential over the medium-term. ### Key macroeconomic statistics (from scenario table and narrative) - Real GDP growth (percentage): Baseline 3.1 (2022), 0.7 (2023), 1.1 (2024), 1.2 (2025), 1.2 (2026); Stress 3.1 (2022), -3.8 (2023), -2.2 (2024), 1.1 (2025), 4.4 (2026). - Unemployment (percentage): Baseline 5.5 (2022), 6.0 (2023), 6.0 (2024), 5.6 (2025), 5.5 (2026); Stress 5.5 (2022), 6.4 (2023), 9.1 (2024), 9.3 (2025), 7.1 (2026). - Inflation (percentage): Baseline 10.3 (2022), 4.7 (2023), 2.1 (2024), 1.7 (2025), 1.8 (2026); Stress 10.3 (2022), 11.4 (2023), 3.9 (2024), 1.1 (2025), 2.6 (2026). - 3 months EURIBOR (percentage, year average): Baseline 2.8 (2022), 3.2 (2023), 2.9 (2024), 2.8 (2025), 2.8 (2026); Stress 2.8 (2022), 6.2 (2023), 4.9 (2024), 3.2 (2025), 2.7 (2026). - House Price Index (2022=100): Baseline 100.0 (2022), 101.6 (2023), 101.4 (2024), 99.9 (2025), 98.3 (2026); Stress 100.0 (2022), 75.6 (2023), 71.0 (2024), 75.8 (2025), 81.7 (2026). - CRE Price Index (2022=100): Baseline 100.0 (2022), 101.6 (2023), 101.4 (2024), 99.9 (2025), 98.3 (2026); Stress 100.0 (2022), 70.9 (2023), 66.3 (2024), 78.2 (2025), 85.6 (2026). - Equity prices (Index, 2022=100): Baseline 100.0 (2022), 100.7 (2023), 101.8 (2024), 103.0 (2025), 104.3 (2026); Stress 100.0 (2022), 71.2 (2023), 66.4 (2024), 74.1 (2025), 83.2 (2026). - 10-year German Sovereign Bond Yield (percentage): Baseline 2.1 (2022), 2.2 (2023), 2.2 (2024), 2.2 (2025), 2.2 (2026); Stress 2.1 (2022), 5.4 (2023), 4.2 (2024), 2.2 (2025), 1.8 (2026). - 10-year Belgian Sovereign Bond Yield (percentage): Baseline 2.9 (2022), 2.9 (2023), 2.9 (2024), 3.0 (2025), 3.0 (2026); Stress 2.9 (2022), 6.1 (2023), 5.0 (2024), 2.9 (2025), 2.5 (2026). - 1-year Belgian Sovereign Bond Yield (percentage): Baseline 3.5 (2022), 3.2 (2023), 3.1 (2024), 3.0 (2025), 2.8 (2026); Stress 3.5 (2022), 6.2 (2023), 5.1 (2024), 3.4 (2025), 2.8 (2026). ### Fiscal outlook - The chapter frames the fiscal outlook as challenging given weakened activity, elevated inflation, and uncertainty about tightening financial conditions. --- ### Financial sector resilience and structure - Banks’ capital adequacy ratio for systemically important institutions (SIs) stood at 18.3 percent at end-2022, dominated by CET1. - Profitability is at pre-pandemic levels and non-performing loans (NPLs) have continued to decline. - Liquidity Coverage Ratio (LCR) is 163 percent and Net Stable Funding Ratio (NSFR) is 128 percent. - Main funding source is retail deposits, with 60 percent insured. - Insurance sector: solvency ratios higher than in the pre-pandemic period. - Investment funds did not face redemption shocks and did not need to test liquidity management tools during the pandemic. Key financial structure metrics - Banks account for 52 percent of financial sector assets at 239 percent of GDP. - Insurance companies account for 59 percent of GDP. - Captive financial institutions (CFIs) were 417 billion as of 2022Q2; their share in the financial system has decreased. - Total assets of pension funds are 40 billion at end-2022. - Outstanding debt securities (mostly government) are approximately 140 percent of GDP. - Market value of listed shares is about 50 percent of GDP. Concentration and supervision - Number of institutions: 30 domestic licensed credit institutions, 46 EEA branches, and five non-EEA branches in Belgium. - The four largest banks account for 73 percent of assets; foreign ownership is 48 percent. - Ten banks with 78 percent market share are identified as Significant Institutions (SIs) and supervised by the Single Supervisory Mechanism (SSM). - Excluding foreign branches, SIs’ market share is 88 percent. - 16 Less Significant Institutions (LSIs), with a 3 percent market share excluding Euroclear Bank, are supervised by the NBB with ECB oversight. - LSIs on average had higher capital adequacy and liquidity ratios than SIs at end-2022; LSIs’ profitability levels are higher compared to SIs. Insurance and investment funds - Large banks offer insurance products; banks’ market share in insurance distribution is 17 percent. - The top ten insurers account for nearly 70 percent of premium income. - Investment funds with characteristics that might make them susceptible to runs have assets of about 28 percent of GDP. - Asset under management (AuM) of Belgian investment funds susceptible to run risks included in the stress testing exercise was approximately 153 EUR bn at end-2022. - Public open-ended investment funds (OEFs) are almost entirely plain UCITS sold to retail investors, with low leverage and limited exposure to derivatives and off-balance sheet activities. --- ### Financial sector vulnerabilities and risks Interest-rate related risks - Rapidly rising interest rates may pose challenges: a considerable part of bank assets are in long-term fixed rate mortgages, causing assets to reprice slowly after ECB rate increases. - Deposit rate increases have remained modest to date, improving net interest margins; pressure to raise deposit remuneration may mount. - Phase-out of the ECB’s favorably priced liquidity facilities in June 2023 increases funding cost pressures. - Available-for-sale and held-to-maturity (HTM) bonds are only a small share of total assets, limiting solvency risks from mark-to-market moves. - Insurers: investment in government bonds 40 percent and corporate bonds 20 percent outside unit-linked business; small duration gaps limit interest rate risk for insurers. - Holdings of loans (including long-duration mortgages) account for around twelve percent of total insurer assets. Household sector risks - Stage 2 loans since year-end 2021 rose from 16.3 percent to 19.1 percent for non-financial corporations (NFC) and from 8.2 percent to 9.9 percent for households by Q12023. - Household debt rose from less than 40 percent to more than 60 percent of GDP over the past two decades as mortgage lending expanded. - Residential mortgages account for a fifth of bank total assets and 55 percent of GDP; insurers hold mortgages equal to 5 percent of investments. - Riskier mortgage segments with elevated LTV and DSTI ratios could be vulnerable if macro-financial conditions deteriorate. - Mitigants: real estate valuations remain lower than in peer economies; fixed-rate and fully amortizing mortgages are widespread; robust labor market, automatic wage indexation, pandemic excess savings, and financial assets far surpassing liabilities. - NBB prudential guidelines since 2020 tightened LTV and DS(T)I limits; additional bank capital required against residential mortgage lending since 2013. Corporate sector risks - Aggregate NFC debt is more than 140 percent of GDP, dominated by company-to-company lending, accounting for nearly 60 percent of the total. - Remaining corporate debt has increased modestly over two decades; credit quality strengthened since end-2019. - Cost pressures from commodity gyrations and automatic wage indexation, along with rising interest rates, have raised financing costs and credit risk concerns for weaker firms. Commercial real estate (CRE) risks - Belgian banks’ loan portfolio is among the most exposed to CRE-collateralized NFC lending within the euro area: CRE exposure is 15 percent of GDP or more than a fourth of total NFC credit. - Nearly three quarters of outstanding CRE exposures show LTV ratios of less than 60 percent; more than a tenth are at 80 percent and beyond. - CRE accounts for 12 percent of insurers’ investment portfolios. - A downturn in CRE could reduce borrower debt-servicing capacity and collateral recoverable values, straining the financial sector. --- ### Scenarios and stress testing Adverse scenario design - The adverse scenario combines global and Belgium-specific shocks: intensifying repercussions of the Ukraine war, exacerbated energy crisis, supply chain disruptions, persistent inflation, subsequent monetary tightening, sharp fluctuations in real interest rates, risk premiums, and asset repricing, and insolvencies in weaker financial institutions causing market dislocations and cross-border spillovers. - The scenario is simulated using the Global Macrofinancial Model (GMM), a structural macro-econometric model of the 40 largest economies. Bank solvency stress test overview - The FSAP solvency stress test is a top-down exercise covering all seven SIs that represent over 90 percent of banking sector assets: five domestic banking groups and two subsidiaries of large foreign banking groups. - The exercise uses the IMF’s solvency stress-testing framework and supervisory data for Q4 2022 provided by the SSM; satellite models estimated using aggregate data from the NBB. - A quasi-static approach is used for growth of banking and trading books; asset allocation and funding composition remain the same; balance sheets grow in line with the scenario nominal GDP path with a floor of zero percent on balance sheet growth to prevent deleveraging. - Credit risk is modeled across household lending, corporate lending, sovereign lending, corporate bonds, financial institution bonds, and sovereign bonds in both banking book (AC) and trading/FVOCI holdings. Credit risk and NPL projection methodology - Transition matrices (TM) are projected via Beta linking when long historical TM time-series are lacking; credit impairments follow IFRS9 calibration. - NPL ratios are projected using an econometric model at quarterly frequency and annualized. Best-fit macrofinancial variables: unemployment, property prices, and real GDP growth. - Aggregate NPL ratios are adjusted to banks and lending types using a Bayesian rule linking banks’ shares of lending exposures and non-performing exposures to sector aggregates. - Equations specified for translating sector-level NPL ratios to bank-level total and mortgage NPL ratios (see equations (1) and (2) in source). Bank solvency scenario table (selected entries repeated for clarity) - Real GDP growth (percentage): Stress -3.8 (2023), -2.2 (2024), 1.1 (2025), 4.4 (2026). - Unemployment (percentage): Stress 6.4 (2023), 9.1 (2024), 9.3 (2025), 7.1 (2026). - Inflation (percentage): Stress 11.4 (2023), 3.9 (2024), 1.1 (2025), 2.6 (2026). - 3 months EURIBOR (percentage, year average): Stress 6.2 (2023), 4.9 (2024), 3.2 (2025), 2.7 (2026). - House Price Index (2022=100): Stress 75.6 (2023), 71.0 (2024), 75.8 (2025), 81.7 (2026). - CRE Price Index (2022=100): Stress 70.9 (2023), 66.3 (2024), 78.2 (2025), 85.6 (2026). - Equity prices (Index, 2022=100): Stress 71.2 (2023), 66.4 (2024), 74.1 (2025), 83.2 (2026). - 10-year Belgian Sovereign Bond Yield (percentage): Stress 6.1 (2023), 5.0 (2024), 2.9 (2025), 2.5 (2026). --- *Source: IMF staff calculations.* ### 19. NPL ratios for other types of lending are estimated in the same way. A separation of ### 19. NPL ratios for other types of lending are estimated in the same way. A separation of ### NPL / PD estimation - NPL ratios for other types of lending are estimated analogously; separation between secured and unsecured NPLs is not available historically or at time 0 and is therefore derived through LGD modelling. - Lending probabilities of default (PD) are derived through the respective NPL ratio projection (Serwa, 2016). - Cross-border household NPL ratio is assumed to follow the same growth path as the domestic NPL ratio. - For small and medium enterprise (SME) lending, PDs are derived through the corporate NPL ratio in the same way. - For large domestic corporate lending and lending to financial institutions, two independent PD paths are projected: - PD path informed by NPL ratio projections. - PD path from a panel model integrating average expected-default-frequency (EDF) data obtained from Moody’s, sensitive to country-specific macro-financial variables (real GDP growth and sovereign spreads). - EDF-based corporate projections cover Belgium and ten other countries: Czech Republic, France, Germany, Italy, Luxembourg, Netherlands, Slovakia, Türkiye, UK, and USA. - The two PD paths (NPL-based PDs and EDF-based PDs) are integrated into a singular PD using linear programming (methodologies referenced: Reeves and Lawrence, 1991; Lam et al., 2001; Panagiotopoulos, 2012). - For lending to financial institutions (domestic and cross-border), PDs are projected via a panel model using the average EDF for financial institutions, with the output gap as the defining macro-financial explanatory variable. - EDF-based PD projections are extended to all debt securities classified as amortized cost (AC) and fair value through other comprehensive income (FVOCI). - Sovereign debt securities PDs are derived from the respective spread paths of sovereign bonds under the baseline and stress scenarios. - Bank-specific adaptation of aggregate PDs: all PD paths not derived from NPL ratios are estimated on aggregate level and adapted to bank specific PDs through formula (3): - PD_it = N(G(NP_i0)) + G(aPD_t) − G(aPD_0) (referenced as formula (3)) - where PD_it is the bank specific probability of default and aPD_t is the aggregate probability of default at time t; G is inverse of cumulative distribution function of standard normal distribution and N is cumulative distribution function of standard normal distribution. ### Loss Given Default (LGD) estimation - LGD for collateralized lending is calibrated through structural modelling using reported loan-to-value (LTV), starting point reported LGDs, and property price paths. - LGD for unsecured lending is calibrated through the Frye-Jacobs method (Frye and Jacobs, 2012). - Both secured and unsecured LGDs are calibrated at the bank-specific portfolio level. - Under the adverse scenario, substantial decline in property values materially increases LGD for secured lending, materially impacting credit risk charges associated with mortgage lending. ### Risk Weighted Assets (RWAs) estimation - Differentiation between standardized (STA) and internal ratings-based (IRB) portfolios in RWA calibration. - RWAs change due to balance sheet growth, new provisions for credit losses, exchange rate movements, and the triggered portion of off-balance sheet items. - For IRB portfolios, the Asymptotic Single Risk Factor (ASRF) model for unexpected losses is implemented (according to Basel III). - Regulatory through-the-cycle (TTC) PDs are calibrated through scenario point-in-time (PiT) projections using a smoothing parameter recommended by the Belgian authorities. - Regulatory downturn (DT) LGD is considered as the maximum between the reported DT LGD at period 0 and the estimated PiT LGD. ### Market Risk methodology - Solvency stress tests assess market risk from interest rates, exchange rates, and equity prices; derivatives portfolio assessment limited by deficiency in access to detailed data. - Market valuation losses for debt securities (sovereigns, financial institutions, large corporates) are estimated using a modified duration approach; current average modified duration provided by the EBA transparency exercise. - Analysis focuses on trading book debt securities measured at fair value (FVPL or FVOCI); HTM portfolio (AC) stressed similarly but unrealized losses from HTM were not integrated into solvency stress test (used in solvency-liquidity interaction analysis). - Debt securities subject to three shocks: yield shocks, FX position, and risk-free rate. - Modified duration updated through formula (4). - Change in value of a security calculated through formulae (5), (6), and (7), where percentΔFV_credit spread and percentΔFV_risk free are percentage changes in FV due to credit spread and risk-free shocks respectively. - Spreads projected using linear autoregression time-series models on aggregate historical debt securities data provided by the NBB; average path used as baseline, stress projection taken as the 85th percentile of simulation results. - Risk-free rate derived from the scenario (10-year Germany bond rate). - Sovereign yield curves constructed by linear interpolation of short- and long-term interest rates as specified in macro scenarios. - Non-sovereign yields move in line with sovereign yield with a credit spread at the three-year horizon. - Debt holding valuations assume 50 percent hedging for interest rate risk and FX. - Market valuation losses for commodity, FX, and equity securities estimated as starting position multiplied by changes in commodity prices, FX, and equity price paths. - Equity revaluation subject to a floor constraint in formula (8): ΔNIE = −0.3 × (NIE_LL + NIE_ShLct). ### Net Interest Income (NII) and IRRBB - Bank interest rates on new business estimated and used as input for interest rate risk in the banking book (IRRBB); time-series regressions based on NBB data project aggregate funding and lending rates on new and variable-rate business. - Historical interest rates for residential mortgages, unsecured consumer lending, and corporate lending are co-integrated and modeled via a Vector Error Correction Model (VECM), with EURIBOR as an exogenous variable. - Single equation time-series estimations model rates for term deposits, retail overnight deposits, corporate overnight deposits, and savings using EURIBOR as independent variable. - Projected funding and lending rates are mapped to banks’ financial assets and liabilities by product and counterparty using the Short-Term Exercise (STE) IRRBB template (provides bank-specific maturity ladder and repricing dates). - Effective funding costs integrate repricing structure; pricing of repos aligns with changes to EURIBOR. - Yields of new debt securities evolve based on scenario projections of the risk-free rate (10-year Germany bond). - Interest income projections apply effective lending rates to performing loans only; weighted average lending rates used with December 2022 notional weights for consumer loans, mortgages, and non-financial corporates. - Derivatives have small proportion relative to total interest-bearing assets/liabilities and do not play pivotal role in adjusting interest income/expense. ### Non-Interest Income and other P&L projections - Non-interest income (non-II) projected via Monte Carlo simulation (10,000 alternative non-II paths); average used for baseline, 10th percentile used for adverse scenario. - Both baseline and stressed non-II projections adapted to nominal GDP growth per balance-sheet growth rule. - Other expenses in P&L and rest Other Comprehensive Income (rOCI) projected as average of last 10 years (excluding 2020), weighted by total assets. - Other expenses are −79 bps of total assets. - rOCI is −10 bps of total assets. - Income tax estimated at fixed rate when profit before tax (PBT) is positive (without counting OCI): tax rate set at 25 percent. - Dividend rules: expected rate of dividends assigned per bank per NBB information; fixed dividend rate distributed if both profit after tax (PAT) and total comprehensive income (TCI) are positive; total dividends cannot exceed TCI of the bank in the same year. ### Results and key projections - Overall resilience: All banks meet Pillar I capital adequacy requirements over the stress testing horizon. - Baseline: aggregate CET1 capital ratio increases from 18.3 to 22.4 percent between 2022–2 6. - Adverse scenario: - Aggregate CET1 capital ratio declines by 4.3 percentage points to 14 percent at end-2026. - Banks record weakened profits in the first year of the scenario (ROA drops from 58bps in year 0 to 28bps in year 1); losses recorded in the remaining three years of the scenario. - All banks meet minimum capital requirements, but one bank does not meet its Capital Conservation Buffer/Other Systemically Important Institutions Buffer (CcoB/O-SIIB). - Credit impairments and provisions: - Four-year cumulative credit impairments are 108.3 percent of starting CET1 capital by end-2026 under the adverse scenario. - Under the baseline, cumulative four-year impairments are 48.9 percent of starting capital. - Most credit risk impairments are recorded during the last three years of the scenario; bulk originates from mortgage lending, exacerbated by considerable shock in property prices. - Net Interest Income dynamics: - Average NII over total assets in the adverse scenario rises to 2.4 percent by end-2023 when the average 3-months EURIBOR in the scenario increases to 6.2 percent. - In the baseline, NII is 2.1 percent at end-2023 with average 3-months EURIBOR during the year at 3.2 percent. - Average annual risk-free rate increases from 2.1 to 5.4 percent in 2023 and drops to 1.8 percent at the end of the stress scenario. - Average NII ratio in the adverse scenario drops to 1.5 percent at end-2025 due to decrease in the base rate while risk-free rate remains high. - Market risk losses: - High in the first year of the adverse scenario, then fall in later years. - Four-year cumulative market losses are 0.9 percent of starting CET1 under stress, versus 0.2 percent in baseline. - Stressed non-interest income lower than baseline but remains positive on average. - Heterogeneity across banks: - Variation in credit provisions/impairments linked to average LTV ratios and lending portfolio composition. - Market losses exhibit more consistent pattern across banks with limited divergence. ### Figures and projected paths (methodology notes) - Satellite models and projections are estimated on a quarterly basis; charts presented at annual frequency for the stress test exercise. - Spreads and EDF projections for multiple countries constructed from aggregate historical data provided by the NBB and Moody’s EDF data, simulated via time-series models; stress paths taken from higher percentiles of simulations. - HTM unrealized losses excluded from solvency stress test but used in solvency-liquidity interaction analysis. ### Sensitivity analysis (scope) - Section H introduced as "Sensitivity Analysis — Scope" (no substantive content in provided extract beyond heading). *Source: IMF staff calculations.* ### 47. The solvency stress test analysis projections indicate a significant increase in NII ### 1belea2023011 - 47. The solvency stress test analysis projections indicate a significant increase in NII ### Solvency stress test: sensitivity analysis design - Two additional scenarios layered on the adverse scenario: the "20 percent transition" and the "50 percent transition". - These scenarios simulate depositor responses to climbing interest rates by shifting non-term deposits into term deposit accounts. - Identical pass-through rates as in the NII projection of the primary solvency exercise are used. - Credit losses and market losses are held the same as in the main stress scenario to isolate the effects of depositor behavior. ### "20 percent transition" scenario — results - Assumption: 20 percent of the bank's non-term deposits migrate to term deposit accounts. - Capital and profitability trajectories over the four-year stress period: - CET1 ratio decreases to 13.1 percent in the final year. - NII over total assets declines from 2.3 percent in year one to 1.4 percent in year four. - Return on assets (ROA) moves from 0.3 percent in the first year to -0.6 percent by the fourth year. - One bank fails to meet minimum CET1 requirements by the fourth year. - Primary driver of the shortfall: substantially higher rates on term deposits in year zero that continue to rise, while non-term deposit rates are essentially null in the initial year and do not rise commensurately during the stress period. ### "50 percent transition" scenario — results - Assumption: 50 percent of non-term deposits transition to term deposit accounts. - Capital and profitability trajectories over the four-year stress period: - CET1 ratio declines to 11.5 percent in the final year. - NII over total assets falls from 2.0 percent in the first year to 1.4 percent in the fourth year. - ROA turns negative from the second year onwards, with values ranging from -0.1 percent to -0.6 percent in the fourth year. - The same bank as in the 20 percent scenario fails to meet minimum CET1 requirements by the fourth year. ### Comparative findings from sensitivity analysis - Both transition scenarios show decreasing profitability and capital adequacy ratios, highlighting the significant impact of depositor behavior changes on asset-liability management. - The analysis underlines the need to incorporate depositor behavior into stress testing exercises to better capture ALM dynamics and supervisory risk assessment. ### Recommendations (from solvency analysis) - Strengthen the General Stress Testing Framework: - Augment the current stress testing framework to better integrate all distinct models and focus on different areas of risk (e .g., credit risk, NII) for a holistic understanding of cumulative impacts on profitability and capital adequacy. - Incorporate IFRS 9 Approach into Credit Risk Modelling: - Integrate an IFRS 9-based approach to introduce a more forward-looking assessment of credit losses. - Utilize Advanced Tools for Assets and Liabilities Management (ALM): - Continue incorporating more advanced tools such as dynamic simulation models and stochastic modelling in top-down stress testing exercises to capture complex behaviours and a broad spectrum of outcomes. ### LCR stress testing exercise — design and key outcomes - Methodology: - Scenarios obtained by recalibrating Basel weights; six stress scenarios simulate mild and severe stress on asset values, inflows, and outflows. - Under “aggressive” scenarios changes correspond roughly to a doubling/halving of the Basel coefficients. - Scenario types include: "Retail", "Wholesale", "Retail & Wholesale", "Market", "Wholesale Total", and "Total Stress". - System-level outcomes: - Starting LCR around 150 percent. - The system remains generally liquid in all scenarios except the most severe one. - Two out of seven banks consistently fall below the regulatory hurdle of 100 percent in most scenarios. - In the most severe scenario only two banks remain above the regulatory hurdle. - Sensitivity to coefficient changes: - System-wide LCR falls below the regulatory threshold when Basel coefficients are doubled. - No banks fall below the regulatory hurdle for mild coefficient changes. - An increasing number of banks exhibit LCR below 100 percent when coefficients are increased by 50 percent. ### Cash flow stress testing exercise — design and key outcomes - Methodology: - Cash flow analysis over a 12-month period under three stress scenarios: (i) Market Stress, (ii) Retail and Wholesale Stress, (iii) Combined Stress. - Tracks contractual cash inflows and outflows across maturity buckets and assesses counterbalancing capacity (CBC). - Runoff and haircut calibrations: - Runoff rates calibrated using similar daily rates as the ECB (2019). - Deposit outflow assumptions: stable retail deposits run off about 10 percent over three months under high stress. - Non-operational deposits from financial and non-financial customers: outflows range between 30 percent (low stress) and 50 percent (high stress). - Haircuts on assets: low stress corresponds to the 95th percentile and high stress to the 99.5th percentile of a lognormal distribution around ECB haircuts. - Computed haircuts range between 7 and 30 percent in the “low stress” scenario and 15 and 40 percent under the “high stress” scenario. - System-level liquidity results: - Even in the most extreme scenario, the system maintains an adequate level of counterbalancing capacity. - Initial counterbalancing capacity approximately 24 percent of total bank assets. - Counterbalancing capacity diminishes by a factor of four after 12 months but the overall system remains liquid. - Bank-level liquidity shortfalls: - Two out of seven banks encounter a liquidity gap mostly due to deposit outflows; liquidity shortages emerge after about two months. - When deposit outflows are assumed to be frontloaded over the first month, the number of banks with shortages does not change, but those banks run out of counterbalancing capacity after only three days. - Currency-specific findings: - US dollars liquidity is considerably weaker for banks with USD exposure. - Four of seven banks have USD exposure; of these, two banks show USD liquidity shortages similar to the aggregate exercise while two have sufficient USD counterbalancing capacity to meet net outflows over 12 months. ### Key numeric highlights - Transition scenario parameters: - "20 percent transition" and "50 percent transition" - "20 percent transition" outcomes: - CET1 ratio: 13.1 percent (final year) - NII over total assets: 2.3 percent (year one) to 1.4 percent (year four) - ROA: 0.3 percent (year one) to -0.6 percent (year four) - "50 percent transition" outcomes: - CET1 ratio: 11.5 percent (final year) - NII over total assets: 2.0 percent (year one) to 1.4 percent (year four) - ROA: -0.1 percent to -0.6 percent (by fourth year) - LCR and liquidity parameters: - Starting LCR: around 150 percent - Regulatory LCR hurdle: 100 percent - Banks analyzed: seven - Initial counterbalancing capacity: approximately 24 percent of total bank assets - Deposit runoff assumptions (cumulative over three months, percent): - Stable retail deposits: 5-10 - Other retail deposits: 20-25 - Operational deposits: 25-40 - Non-operational deposits from fin. customers: 30-50 - Non-operational deposits from non-fin. & other customers: 30-50 - Committed credit facilities: 60-12 - Liquidity facilities: 75-75 - Haircut ranges on counterbalancing capacity: - Low stress: between 7 and 30 percent - High stress: between 15 and 40 percent - Haircut percentiles used: 95% (low stress), 99.50% (high stress) - Sensitivity tests: doubling Basel coefficients; increasing coefficients by 50 percent *Source: IMF staff calculations.* ### 68. The qualitative NSFR test reveals the heavy reliance of banks stable funding sources, ### 1belea2023011 - 68. The qualitative NSFR test reveals the heavy reliance of banks stable funding sources, ### Bank funding structure and liquidity measures - Deposits from household and non-financial corporations made up more than half of total banks liabilities by end-2022. - Debt securities (excluding CDs) were about 6 percent of total liabilities. - Net stable funding ratio (NSFR) at the end of 2022 stood at 132 percent. - Funding structure components (as presented): Deposits from Non-Financial Corporations; Deposits from Households; Deposits from Central Bank and General Government; Deposits from Banks and Financial Institutions; Certificates of Deposits; Covered Bonds and Other Debt Securities; Derivatives; Other Liabilities. - Counterbalancing capacity and liquid asset categories shown include: Coins and Banknotes; Central Bank Reserves; Level 1 Tradable Assets; Level 2 Tradable Assets; Other Tradable Assets; Non-Tradable Assets; Committed Facilities. ### Solvency–liquidity interaction: scope and methodology - Objective: explore solvency-liquidity risk interplay under stress, focusing on whether banks can withstand losses from forced liquidation of HTM portfolio when counterbalancing capacity is exhausted. - HTM portfolio characteristics: - HTM is 7.8 percent of bank assets (approximately 73 percent of total debt securities held by the banking sector). - HTM composition: sovereign debt 65.5 percent; debt from credit institutions 20.6 percent; corporate debt 7.4 percent; non-bank financial institutions 6.1 percent; central banks 0.4 percent. - Forced liquidation assumptions: - Banks sell HTM assets at market value (devaluation vs. book value) only after exhausting tradable counterbalancing capacity. - Devaluation trajectory follows the risk-free rate in the adverse scenario from 2022 through end of either year 1 or year 3 of the scenario (variant-dependent). - No additional haircuts assumed; HTM primarily HQLA (mainly low yield sovereign bonds). - Losses equal the market-book value gap for the fraction sold; losses hit PnL. - RWAs corresponding to total utilized counterbalancing capacity (not only HTM sales) are removed from balance sheets. - Provisions held against sold HTM are reversed and recognized as income in PnL. - Four forced-liquidation scenarios (standalone events): - Scenario 1: "Market" — 47.4 percent of HTM sold (aggregate). - Scenario 2: "Outflow" — 66.6 percent of HTM sold (aggregate). - Scenario 3: "Combined" — 86.6 percent of HTM sold (aggregate). - Scenario 4: "Complete liquidation" — 100 percent of HTM sold. - Analysis constrained to the year of forced liquidation; no modeling of multi-year deposit outflow effects on net interest income. ### Solvency–liquidity interaction: results (Year 1 and Year 3) - Year 1 (pre-liquidation position): - Average CET1 ratio: 18.3 percent. - Return on assets: 28bps. - Total portfolio devaluation under forced liquidation: 7 percent. - Losses as proportion of CET1 (before provisions reversal): - Market: 453bps - Outflow: 636bps - Combined: 826bps - Complete liquidation: 955bps - Reversal of provisions (gains to PnL) across scenarios: 31bps to 95bps. - Net losses over pre-liquidation CET1 after provisions reversal: - Market: 421bps - Outflow: 626bps - Combined: 766bps - Complete liquidation: 860bps - RWAs reduction due to forced liquidation: 2 percent. - Post-liquidation CET1 ratios (after 2 percent RWA reduction): - Market: 18.4 percent - Outflow: 18.3 percent - Combined: 18.2 percent - Complete liquidation: 18.1 percent - Outcome: all banks remained above minimum capital requirements. - Year 3 (pre-liquidation position): - Average CET1 ratio: 15.4 percent. - Return on assets: -65bps. - Total portfolio devaluation under forced liquidation: 4.8 percent. - Losses as proportion of CET1 (before provisions reversal): - Market: 364bps - Outflow: 511bps - Combined: 663bps - Complete liquidation: 766bps - Reversal of provisions (gains to PnL) across scenarios: 37bps to 111bps. - Net losses over pre-liquidation CET1 after provisions reversal: - Market: 328bps - Outflow: 500bps - Combined: 592bps - Complete liquidation: 655bps - RWAs reduction due to forced liquidation: 2 percent. - Post-liquidation CET1 ratios (after 2 percent RWA reduction): - Market: 15.3 percent - Outflow: 15.2 percent - Combined: 15.1 percent - Complete liquidation: 15.0 percent - Post-liquidation return on assets in scenarios: -78bps, -82bps, -83bps, and -85bps respectively. - Outcome: all banks continued to meet minimum capital requirements. - Overall inference: - Forced liquidation can impose significant losses, but the 2 percent RWA reduction and reversal of provisions partially offset these losses. - CET1 ratios decline and return on assets deteriorate, yet capital levels remain robust enough for minimum requirements. - The exercise is presented as a complementary, immediate-impact tool alongside longer-term solvency and cash-flow stress tests. ### Insurance stress test: scope, scenarios, and modeling assumptions - Coverage and sample: - Eight composite insurers included: KBC, Belins, Ethias, Axa, Baloise, AG Insurance, Allianz, and P&V — covering around 76 percent of the Belgium insurance market. - Six firms participated in the bottom-up exercise (all except Allianz and P&V). - Aggregated balance sheet assets of participating insurers: €231 billion. - One firm uses a full internal model; one uses a partial internal model. - All assessed insurers apply the volatility adjustment (VA); no other LTG measure applied among the sample. - Investment profile and risks: - Investments dominated by fixed income assets and significant direct and indirect real estate. - More than two thirds of fixed income investments are government bonds. - 87 percent of government bond investments have an A or higher rating. - Corporate bond investments: more than half of all investments have a credit rating of A or higher; financials represent 38 percent of total corporate bond investments. - Corporate high yield exposures cover less than three percent, but could increase with rating migrations due to large share of BBB investments. - Cash levels: above one percent of total assets. - Mortgages (part of real estate investments) have weighted average LTV of 74 and DSTI of 34 percent. - Home bias: around half of government bonds held are Belgian sovereign; France, Germany, and Spain cover another quarter. - Solvency position: - Market risk accounts for about 70 percent of the basic Solvency Capital Requirements (SCR); main drivers are spread risk, property risk and equity risk. - Duration gaps small; interest rate risk contribution to capital requirements relatively small. - Unrestricted tier 1 capital represents around 80 percent of insurers’ eligible own funds. - Unrestricted tier 1 capital alone covers the SCR, though some insurers’ reliance on tier 2 and 3 capital is close to regulatory limits. - Scenario design: - Macrofinancial scenarios from the banking stress test used as basis; shocks front-loaded (instantaneous shock at reference date 31 December 2022). - Risk-free rate term structure adjusted to reflect significant downward reversal after 2023 in the scenario. - Market risk shocks include: risk-free interest rate, equity and property prices, default rates of mortgage loans, credit spreads of corporate and sovereign bonds. - Increase in credit spreads raises the VA (per Solvency II method), which increases the effective risk-free rate and offsets much of the credit spread shock for insurers applying VA. - Bottom-up exercise included a (30 percent) mass lapse shock on non-mandatory in-force life policies (inclusive of normal lapsation) and an inflation shock following the scenario CPI and wage paths; CPI assumed to return to two-percent by 2032 and remain constant thereafter; wage inflation assumed to increase at the same pace as CPI after 2027. - Capital measure and valuation: - Solvency II mark-to-market framework applied, with LTG measures (mainly VA) affecting discounting of insurance liabilities. - Main output: effect on own funds eligible for SCR coverage; SCR partially recalculated after stress. - Bottom-up results provided more granular assessment of own funds composition. *Source: IMF staff calculations; NBB.* ### 90. To benchmark the results of the bottom-up stress test and assess sensitivities to ### 1belea2023011 - 90. To benchmark the results of the bottom-up stress test and assess sensitivities to ### Top-down (TD) stress test design and scope - TD stress test run by the FSAP team to benchmark bottom-up (BU) results and assess shock sensitivities. - Analysis based on data received from the NBB and insurance undertakings, including: - granular breakdown of investment assets, government and corporate bond holdings, maturities and coupon rates of fixed income investments; - detailed maturity breakdown on cash flow projections. - Focus predominantly on the impact of financial shocks: rating migrations of insurers’ corporate bond portfolio, exposure to the Belgium sovereign, and real estate exposure. ### Data sources and reporting templates used - Data gathered in line with Solvency II quantitative reporting templates and national-specific templates. - Solvency II templates used for solvency and liquidity assessments: - Balance sheet (S.02.01) - Open derivatives (S.08.01) - Securities lending and repos (S.10.01) - Asset-by-asset investment holdings (S.06.02) - Cash-flow projections (S.13.01, S.18.01) - Impact of LTG measures and transitionals (S.22.01) - Own funds (S.23.01) - Calculation of the Solvency Capital Requirement (S.25.01 – S.25.03, S.26.01) - National templates included NBB “liquidity reporting” and collection of the standard formula calculation of the SCR by full and partial internal model users. ### TD methodology and modelling choices - Shocks from the scenario applied to investment assets and insurance liabilities: - Haircuts applied to market values of assets per scenario. - Revaluation of fixed income assets with the stressed term structure. - Technical provisions (except for unit linked business) approximated with the stressed term structure including the volatility adjustment (VA). - Re-calculation of SCR limited to selected risk modules within market risk component: - Capital charges for equity risk, spread risk and property risk proportionately adjusted in line with change in exposures due to stress. - Equity risk capital charge corrected for the symmetric adjustment, which changes to -10 percentage points after the fall in equity prices. - All other components of the basic SCR not modelled by TD but derived from the BU exercise. - Reactive management actions are not modelled in TD; they are captured in the BU exercise (insurers asked to show results with and without reactive management actions and to provide details on embedded management actions). ### Results of the Solvency Stress Test — aggregate outcomes - Median solvency ratio: - Drops from 192 percent before stress to 113 percent after stress (BU calculations). - Including reactive management actions, the new median solvency ratio is 142 percent. - Aggregated solvency surplus: - €9.1 billion before stress - Falls to €1.8 billion after stress - €3.3 billion when reactive management actions are included. - VA effect: - Insurers benefit substantially from assumed increase in the VA in the stress scenario; VA mitigates short-term bond spread volatility and immunizes insurers against spikes not driven by underlying credit deterioration. - Without application of VA, most insurers’ solvency ratio would have fallen below 100 percent. - Combined shocks (valuation losses + severe mass lapse) impact: - Solvency ratio fell almost 80 percentage points from a baseline of almost 200 percent (whole sample). - Eligible own funds (EOF) fell by 43 percent. - SCR falls by around seven percent due to lower asset and liability exposures post stress. - Assets over liabilities ratio falls from 110 percent to 107 percent, reducing assets over liabilities (eAoL) by more than eight billion euro, almost halving the year-end 2022 figure. - One insurer would fall below a solvency ratio of 100 percent before the application of reactive management actions. ### Key vulnerabilities and sensitivity analyses - Corporate bond exposure: - Corporate bond holdings represent 15 percent of total assets and are almost entirely investment-grade. - A significant share is held in BBB grade; a downgrade leads to significant valuation losses due to higher spreads. - A rating migration scenario (like the 2008 experience) combined with assumed increases in bond spreads would lead to an average reduction of the solvency ratio of 14 percentage points. - Although SCR can increase as assets move into non-investment grade, this is partially offset by lower exposure following revaluation; solvency ratio change driven mainly by change in own funds. - Belgian sovereign exposure: - Belgian sovereigns represent 18 percent of the balance sheet or 50 percent of the government bond exposure. - Assuming an increase of only 30 basis points (i.e., the difference between the shock on Belgian sovereign and the shock for non-AAA or AA rated EA countries in the model specifications), the solvency ratio would see a further decline of nine percentage points or a loss in valuation of around one billion euros. - This is in addition to the original spread shock on government bonds, which reduces the sample’s solvency ratio by 28 percentage points. - In line with Solvency II standard formula, no capital requirements for government bonds have been assumed. - Real estate exposure: - Real estate investments represent about ten percent of Belgian insurers’ balance sheets (direct property, real estate funds, loans and mortgages, bonds, equity of real estate related corporations). - Property (excluding own use) and mortgages and loans are main real estate exposures; real estate related equity investments are significant and can be the dominant equity exposure for several insurers (a third of total equity exposure for the total sample, with a maximum of almost three quarters). - Real estate funds represent nine percent of total fund investments. - Six percent of corporate bonds are allocated to real estate corporations. - Stressing all real estate related exposures leads to, all else equal and with considerable dispersion across insurers, EOF (proxy: excess of assets over liabilities) falls by 19 percent and leads to a similar impact on solvency ratio. - Inflation shock: - Impact small because end 2022 balance sheets already reflect insurers’ own projections of the inflation trajectory. - Inflation shock increases non-life best estimates liabilities by around five percent, negligible for life business. - Main affected line: long-tailed workers compensation insurance, which accounts for six percent of the non-life business. - Some insurers benefit from increase in value of inflation-linked bond holdings, partially offsetting increases in best estimate liabilities. - Part of inflation shock impact is implicitly assumed in the interest rate shock; baseline cashflow projections already incorporate insurers’ own expectations about inflation. ### Management actions, capital structure concerns, and EPIFP - Embedded and reactive management actions: - Have a significant impact on solvency ratios. - Reactive management actions mitigate the impact of the stress scenario on the solvency ratio by 20 percentage points; overall solvency ratio after reactive management action would be at 140 percent (note: not all insurers applied reactive management actions). - Typical actions: dividend retentions, hedging derivatives, intra-group transactions, asset sales, liability de-risking, reinsurance changes, withholding profits, forward bonds and spread locks, portfolio reallocation, intra-group capital management. - Tiering limits and EPIFP: - Tier 1 capital generally sufficient to cover the SCR, but tiering limits can become a constraint under stress. - In baseline, some insurers’ eligible tier 2 and tier 3 capital is close to Solvency II limits: - Sum of eligible tier 2 and tier 3 capital can only cover less than 50% of the SCR. - Eligible tier 3 capital needs to be less than 15% of SCR. - Restricted tier 1 capital can only cover 20% of the SCR. - Expected profits in future premiums (EPIFP) represent a significant share of unrestricted tier 1 capital for several insurers and thus cover a large share of the SCR, raising concerns about loss absorbency and the adequacy of contract boundary application in technical provisions. ### Liquidity risks: mass lapses and variation margin calls - Liquidity analysis focused on: - Liquidity risk from mass lapses (policyholder behavior in response to income shocks, inflation, unemployment, or more attractive rates elsewhere). - Liquidity risk from variation margin calls on interest rate swaps (IRS). - IRS exposure and variation margin mechanics: - Derivatives, especially IRS, are important for duration extension and liability matching; using IRS (floating rate payer, fixed rate receiver) reduces exposure to falls in interest rates but creates vulnerability to increases. - When market value of derivatives falls, insurers required to pay variation margins; in a long position they receive margin payments when interest rates increase, allowing netting where transactions are with the same counterparty. - IRS concentration and maturities: - IRS exposure concentrated among very few insurers with significant holdings. - Those insurers hold IRS in both directions and adjust positions dynamically; high number of off-setting transactions implies small number of counterparties enabling netting. - Weighted average remaining time to maturity: - 12.6 years for floating for fixed IRS - 13.8 years for fixed for floating IRS - Number of open positions: - 181 (floating for fixed) - 519 (fixed for floating) - Most derivatives traded over the counter and not centrally cleared in Belgium. - Data for liquidity analysis: - Data gathered from NBB in line with Solvency II templates: S.08.01 (derivative positions) and S.06.02 (asset by asset investment holdings). - Assessment focused on available liquid assets in stress, excluding assets pledged as collateral. - Variation margin stress scenarios and assumptions: - Very severe interest rate increases of up to 100 basis points within a five-day period modeled. - Three interest rate shocks over two time horizons: - 25 basis point (bp) increase overnight: to meet cash margin calls only the most liquid assets (narrowest definition: cash deposits) could be drawn upon. - 50 bps and 100 bps increases unfolding over five days: insurers could liquidate some of their most liquid high-quality assets; margin calls could be met with cash deposits plus unencumbered sovereign bonds with high credit quality rating (AAA to AA), revalued after the interest rate increase. - Asset encumbrance and repo access: - Encumbrance levels of high-quality assets generally low but vary significantly across insurers; for most, pledged government bonds account for ten percent or less of their total government bond exposure, but some insurers have considerably higher encumbrance. - Encumbered assets mostly comprise Belgian government bonds (two thirds of pledged sovereign bonds). - Only a small share of corporate bonds is encumbered. - Liquidity stress test findings: - Variation margin calls from sudden strong interest rate increases are a manageable exposure for the Belgian insurance sector overall. - Low exposures for most insurers make the sector largely resilient to margin calls following steep interest rate hikes. - Cash buffers vary significantly; for some insurers with larger exposures, cash may be insufficient under a narrow cash definition. - For two insurers, cash may not be sufficient to service margin calls and they may need to engage in forced liquidation of other liquid assets, such as high-quality government bonds, in the prescribed scenarios. - Insurers have access to repo facilities and money market funds, but these sources may be unavailable in a crisis. - Overall, other liquid assets are sufficient if the interest rate increase unfolds over a longer period. - Quantitative liquidity example from stress testing: cash margin calls on interest rate swaps following a 25-bps interest rate increase could be met by drawing on 22 percent of cash equivalents; it would require 82 percent if interest rates were to increase by more than 100bps. *IMF Staff calculations and analysis based on company submissions and NBB data as presented in the source content.* ### 111. We assess the insurers’ robustness against different lapse scenarios given a stressed ### 1belea2023011 - 111. We assess the insurers’ robustness against different lapse scenarios given a stressed ### Lapse scenarios and stress assumptions - Four lapse scenarios applied: - A 30 percent lapsation event for all lapsable contracts (in line with the solvency stress test). - Two scenarios involving only “high-risk contracts” (contracts without fiscal or contractual penalties): - Complete wipe-out of such products. - A 30 percent lapse scenario for high-risk products. - Combination scenario: a 10 percent withdrawal assumption for products with some form of penalty and a 30 percent lapse scenario for high-risk products. - High-risk contracts account, on average, for about 14 percent of lapsable surrender values, with a range between four and 39 percent across insurers. - Market stresses affecting insurers’ liquid assets are assumed to take place before the lapsations. - A parallel shift of 25, 50, and 100 basis points for the risk-free term structure is assumed for financial shocks. - Note on timing: policyholder pay-outs generally occur one month after demand for lapse for individual contracts and two to three months for occupational contracts. ### Liquid asset treatment, haircuts, and included instruments - Deferred pay-outs permit inclusion of a broader set of liquid, non-pledged assets subject to haircuts after applying financial stress assumptions. - Liquid asset categories included (in addition to cash and sovereign bonds): - High quality corporate bonds (excluding financials). - Covered bonds. - Listed equities of non-financials. - Money market, equity, and debt funds. - Haircuts increase with asset riskiness and are broadly aligned with the banking sector LCR. ### Results: insurers’ liquidity capacity and dispersion - Overall finding: insurers have sufficient liquid funds to withstand significant redemptions, but there are high levels of dispersion across firms. - Cash plus sovereign bonds generally suffice to service pay-outs even in stressed market conditions. - Including all liquid assets: - Most insurers would be able to withstand a complete wipe-out of all products without penalties. - On average, insurers could withstand a further surrender of two thirds of their contracts with penalties. - Observations: - Some companies with high levels of products without penalties also have comparatively lower liquid to total asset ratios. - Some insurers more vulnerable to lapse risk also have higher exposure to variation margin calls. - Analysis performed on a stock basis; investment income can mitigate the liquidity shock. ### Investment income and cash-flow mitigation - Around seven percent of their fixed income assets mature within a year. - Assuming a quarter of these assets maturing within the first trimester provides a cash inflow of around € 2 billion. - Insurers have an average coupon rate of around three percent on their fixed income investments. - Assuming coupon timing and that assets are only sold after coupon payment would add another € 900 million. - These investment inflows represent: - About 20 percent of the contracts without penalty. - 2.5 percent of total surrender values. - Net cash flows from premiums were assumed to be zero (based on quarterly claims and expense ratios for non-life and life business). - The € 2 billion figure is based on the eight institutions included in the solvency stress test for which detailed asset data were collected. ### Recommendations for insurance supervision (NBB) - Implement liquidity stress tests and liquidity scenarios to identify potential sources of stress. - Supervisors should challenge assumptions insurers apply to availability of liquid assets in a stressed environment, including what is considered as cash equivalent. - Collect surrender/lapse data by type of insurance product to support internal analyses. - Engage with industry to reduce dependence on lower tier capital, as: - Some insurers depend on lower quality capital instruments. - These instruments can become binding in a stressed environment and limit reactive management actions. - While tier 1 capital generally suffices to cover the SCR, tiering limits can become binding in stress. - Run a detailed assessment of the impact of the use of the VA on the solvency ratio in a changing interest rate environment: - Quantitative reporting templates show the VA has a profound impact on the solvency ratio. - The NBB previously assessed that VA impact did not distort the level playing field in the low yield environment. - NBB should assess whether solvency ratio improvement due to the VA can lead to an unlevel playing field across undertakings. - Continue to collect figures on the impact of the VA on the solvency ratio. ### Investment Funds Stress Test — Objective and scope - Objectives: - (i) assess ability of investment funds to withstand severe but plausible redemption shocks; - (ii) identify types of funds potentially more vulnerable to liquidity risk; - (iii) estimate sector’s capacity to transmit shocks to the rest of the financial system. - Stress test does not consider the use of liquidity management tools (LMTs). - Focus on funds susceptible to run: funds of funds (FoFs), mixed vehicles, and fixed income funds. - Sample (end-2022): 435 funds for a total NAV of around €136 billion. - Fund category composition (in percent of total AuM): Bond 8%, Mixed (FoF) 58%, Non-public AIF 12%, Other 22%. - Ownership of investment funds (in percent of NAV): Households hold more than 55 percent of public open-ended vehicles exposed to liquidity transformation risk. ### Methodology for investment funds stress test - Three components: - Calibration of redemption shock (historical approach and heterogeneity approach). - Composition of asset sales (liquidation strategies: pro-rata vertical slicing and waterfall horizontal slicing). - Price impact of sales (based on market depth: ratio of average daily trading volumes to asset volatility), measured under normal and stressed trading conditions. - Redemption shocks calibrated using historical data: - Homogeneity assumption: calibrated based on average of worst 3 percent net flows observed by funds in each category. - Resulting levels of redemptions: - Mixed-FoFs: between 4 percent and 9 percent of NAV. - Non-public alternative investment funds: between 14 percent and 18 percent of NAV. - Redemption Coverage Ratio (RCR) defined as Highly Liquid Assets / Redemption shock. - An RCR greater than 1 indicates enough liquidity buffers to weather redemption shocks; RCR less than 1 indicates the fund would have to sell less liquid assets. - Highly liquid assets estimated at fund-level using portfolio composition and liquidity weights derived from Basel III HQLA. ### Investment funds results and key statistics - Portfolio composition: - FoFs hold more than 40 percent of total NAV and are largely invested in other CIUs including ETFs domiciled in other European jurisdictions. - Unencumbered cash overall amounts to 2 percent of total assets under management. - Exposure to fixed income instruments: overall €34 billion, equally split between sovereign and corporate bonds. - More than 80 percent of corporate bond exposure is to investment grade instruments. - HQLA levels: - Bond funds and mixed investment vehicles: HQLA between 40 percent and 60 percent. - Non-public alternative investment funds: weighted average level of liquid assets around 30 percent. - Resilience overview: - The sector would be able to withstand severe but plausible redemption shocks, but a few non-public alternative funds could run into difficulties. - Less than 2 percent of funds (6.7 percent of NAV) analyzed would not have enough HQLA to meet redemption requests in a market stress and thus present liquidity shortfalls. - Table 8 results (Historical Approach, ES 3%): - Homogeneity (ES 3%): - Mixed (FoFs): 1 fund with RCR < 1; % Funds with RCR < 1: >1%; % NAV with RCR < 1: 0.0% - Non-public alternative funds: 5 funds with RCR < 1; % Funds with RCR < 1: 13%; % NAV with RCR < 1: 46.3% - Total: 6 funds; 1.1% of funds; 5.4% of NAV with RCR < 1 - Heterogeneity (ES 3%): - Mixed (FoFs): 1 fund with RCR < 1; % Funds with RCR < 1: >1%; % NAV with RCR < 1: 0.1% - Non-public alternative funds: 6 funds with RCR < 1; % Funds with RCR < 1: 16%; % NAV with RCR < 1: 57.1% - Total: 7 funds; 1.6% of funds; 6.7% of NAV with RCR < 1 - Vulnerabilities identified: - Portfolios concentrated in other non-domestic CIUs or lacking sufficient diversification. - Mixed funds with potential shortfalls might activate LMTs; non-public alternative funds sold to professional investors may not have that possibility. - Underlying non-domestic funds may lack deep liquidity and may engage in leveraged activities, exacerbating stress transmission. *Source: IMF Staff calculations based on NBB, FSMA, Lipper data and internal stress test frameworks provided in the source content.* ### 134. The redemption shock does not result in significant market price impact. In the ### 1belea2023011 - 134. The redemption shock does not result in significant market price impact. In the ### Liquidity shock and redemption impact - In the reference scenarios, the liquidation of domestic sovereign bonds would amount to 10-13 percent of the assets liquidated by investment funds. - Benefitting from a sufficiently large market depth, under the different liquidation strategies, there would not be any relevant price impact from the funds’ sales. ### Market-based contagion and volatility spillovers - Analysis uses the Diebold and Yilmaz (2009, 2012) connectedness framework applied to equity, bond, and mixed fund indices built from funds’ weekly performances. - Belgian investment funds are more vulnerable to volatility shocks originating in other European jurisdictions and global markets than to spillovers from the domestic market. - Market-based contagion between domestic funds is overall limited to equity funds; mixed vehicles exhibit stronger co-movements and similar inward spillovers from foreign markets. ### Key statistics and model parameters (interbank / contagion) - Interbank contagion methodology: Espinoza and Solé (2010) framework using large exposures data for nine banks (COREP) reported by end of December 2022. - Two spillover scenarios: i) credit shock; ii) credit and funding shock. - Model parameters (domestic interbank analysis): - Lambda (credit shock) = 65 percent. - Delta (funding shock) = 50 percent. - Rho (funding shock) = 35 percent. - Finding: The failure of a single domestic bank would not trigger the failure of another bank; no bank falls below its regulatory minimum capital requirement after shocks to one/several interbank exposures. Vulnerability to spillovers remains low overall. ### Cross-border exposures and contagion - Belgian banks’ foreign claims equal 50 percent of GDP. - Cross-border exposures concentrate over 60 percent towards: Czech Republic, the United Kingdom, the Netherlands, France, and Slovakia, primarily to non-financial sectors. - Domestic exposures concentrated in household, government, and NFC sectors: 61 percent of total domestic exposures. - Cross-border contagion analysis uses BIS consolidated banking statistics for 2022Q4 plus NBB data for Belgian subsidiaries of two institutions headquartered outside Belgium. - Cross-border model parameters: - Lambda = 65 percent. - Delta = 50 percent. - Rho = 45 percent. - Three spillover scenarios tested: i) credit shock to cross-border all sectors; ii) credit shock to cross-border financial sectors; iii) credit and funding shock to cross-border financial sectors. - Simulation result: Belgium is highly vulnerable to contagion from an external shock. Default of all exposures in the Czech Republic, Netherlands, France, Luxembourg, and the UK (aside from the US) would have significant adverse impact on Belgian banks' capital. - Impact from defaults in individual foreign financial institutions is expected to be marginal due to limited Belgian cross-border exposure to financial sectors. ### Cross-sectoral contagion via market interconnectedness - Model analyzes second-round effects of solvency stress tests of banks and insurers through crossholdings of financial instruments. - Insurers’ exposures: about 1 billion Euros to banks’ debt and about €17 billion to investment funds’ issued shares. - Satellite models for yields and share prices estimated for banks and insurance companies; coverage varies by model (e.g., yields on bank-issued debt: unbalanced panel covering six banks; banks’ share prices: three banks; insurers’ bond yields and share prices: three insurance groups). - Revaluation loss computations: - Debt holdings: change in yield × exposure × modified duration. - Share holdings: predicted rate of change of share prices × exposure. - Assumed average durations: - Debt issued by banks: slightly below 4 years. - Debt issued by insurers: about 10 years. - Iterative mapping of equity declines to yields and share prices until convergence. ### Market interconnectedness results - Algorithm converges after about 12 iterations. - Losses from crossholdings are limited: - Banks’ equity to asset ratio declines from 4.267 percent to 4.266 percent. - Insurers’ equity to asset ratio declines from 5.913 percent to 5.777 percent. - Larger decline for insurers explained by exposure to revaluation losses from holdings of bank-issued debt. - Contagion channel from investment funds to insurers is weak because investment funds have very little exposure to bank-issued debt. ### Supervisory and policy recommendations - NBB and FSMA should establish a formal agreement for sharing investment funds data: - FSMA collects a broad set of supervisory data; this should be complemented with investment funds’ portfolios collected by the NBB. - FSMA should include the enhanced information set in their monitoring tool to improve data-driven risk-based supervision. - FSMA should develop and adapt their stress test framework for conducting market-wide liquidity risk analysis: - Use detailed information collected by the NBB together with other supervisory data to assess structural vulnerabilities in the sector. - This assessment can form part of the systemic risk assessment of the NBB. - Additional FSAP recommendations based on investment fund liquidity analysis: - FSMA should monitor more closely Belgian managers of non-public open-ended AIFs to enhance liquidity management, including recommending the availability and use of LMTs where necessary and propose possible amendments to legislation considering the European framework. - FSMA and NBB should continue common reporting on asset management & NBFI, including cooperation to monitor relevant transfers between credit institutions and investment funds managed by affiliated managers. - Given its advanced framework regarding LMTs, FSMA should monitor and assess use of LMTs by IFs within the European framework and consider issuing guidance on their use, taking into account international developments. *Sources: IMF staff calculations.* ### Appendix II. Stress Test Matrix (STeM) ### Appendix II. Stress Test Matrix (STeM) ### Banking Sector: Solvency Stress Test — Top-down by IMF - Institutional perimeter - Exercise: Top-Down by FSAP team. - Institutions included: Seven banks subcategorized as SIs. - Among the SIs, five are domestic and two are subsidiaries of foreign institutions. - All banks are domestically focused, but they have large cross-border exposures. - Market share: Total coverage is about 89.4 percent of the banking sector. - Data and baseline date: - Multiple data vintages: December 2022 (year end, starting point for PnL), and March 2023 (starting point for balance sheet and capital). - Supervisory data: Bank balance sheet and supervisory statistics (including FINREP and COREP), information on interest rate risk in the banking book (IRRBB), liquidity risk and market risk sensitivities (including STE templates) provided by the authorities and the ECB. - Expected Default Frequency sourced from Moody’s. - Further supervisory information including probability of defaults by credit portfolios, and a bank-specific stage transition matrix by portfolio from FINREP. - Market and publicly available data, such as ECB statistical data warehouse on funding and lending rates by type of asset and funding portfolios. - Scope of consolidation: banking activities of the consolidated banking group for banks having their headquarters in Belgium. Foreign subsidiaries are assessed on the unconsolidated level covering domestic activities only. - Coverage of sovereign and non-sovereign securities exposures: debt securities measured through fair value (FVPL and FVOCI) and amortized cost (AC) account. - Coverage of lending exposure: credit institutions, nonbank financial institutions, household, and corporate. - Channels of risk propagation / Methodology - FSAP team satellite models and methodologies. - Balance-sheet regulatory approach. - Market risk treated as an add-on component, with separate calibration; impacts both capital resources (via profit and loss or Other Comprehensive Income (OCI)) and capital requirements (RWA). Impact covers trading book and fair valued banking book items and evolves with balance sheet assumptions. - Traded risk impact from revaluation of trading assets (FVPL) and FVOCI securities by counterparty: central government (by country issuers), credit institutions, other financial institutions, and nonfinancial corporates. - Credit spreads on sovereign, credit institutions and corporate securities interpolated using bank-specific residual maturity at the book and issuer level (sovereign issuers by country and individual corporate issuers by ISIN codes). Credit spreads on other securities estimated on a hypothetical portfolio using a duration proxy. Valuation effects assessed using a modified duration approach. Hedges considered ineffective under stress. - Losses for securities portfolios based on duration approach. Losses on equities (both long and short position) based on stock market price movement specified by the scenario. - For internally modelled exposures (IRB): projection of PiT and TTC PDs, LGD, EAD and RWA. For STA exposures: projection of new flows of defaulted exposures, coverage ratio for defaulted loans, and risk weight downgrade for performing exposures. Corporate PDs for largest exposures proxied by Moody’s EDFs. - Provisioning for IRB and STA modeled using IFRS9 transition matrix approach. Transition matrices, PiT PDs, PiT LGDs for loan and securities classified under AC and FVOCI modeled using COREP data. - Net interest income projection incorporates maturity profile of assets and liabilities. Effective interest rates projected through econometric satellite model where exogenous variables reflect the interest rate environment. Sensitivity analysis examines impact of changes in depositor behavior. - Funding costs projected at the portfolio level using funding structure by product and maturity bucket. Funding projections capture systemic risk (linked to the scenario) and idiosyncratic risk (spreads on debt instruments). Funding cost projections utilize bank level data from COREP templates. Lending rates projected at the system level and attached to bank-specific interest rates and outstanding amount at cut-off date. - Change in RWAs estimated on banking book exposures (credit risk charges – CRC) and market risk exposures (market risk charges – MRC) according to Basel III rules. Additional regulatory risk charges (operational risk charges – ORC and counterparty credit risk charges – CCRC) change according to overall balance-sheet growth assumptions. Balance sheet will follow nominal GDP growth when positive, or remain stable when negative. - Stress test horizon - 2023 Q1–2026 Q4 (4 years) - Tail shocks / Scenarios - Two Scenarios: - A baseline scenario based on the April 2023 WEO macroeconomic projections. - An adverse scenario that captures the key risks in the RAM. This scenario relies on GFM, a structural macroeconometric model of the world economy disaggregated into forty national economies (documented in Vitek (2018)). Scenarios for foreign countries where Belgium has significant exposure are extracted from GFM and internally consistent with country scenarios of other ongoing FSAPs. - Risks and buffers; behavioral adjustments - Risks covered: credit (on loans and debt securities), market (valuation impact of debt instruments through repricing and credit spread risk as well as the P&L impact of net open positions in market risk factors such as foreign exchange risks) and interest rate risk (IRRBB) on the banking book. - Concentration risk via sensitivity analysis. - Solvency and liquidity risk interactions, mainly through asset haircut. - Behavioral adjustment: quasi-static approach for balance sheet growth. Asset allocation and funding composition remain the same; balance sheet grows in line with nominal GDP paths of major geographical exposures and subject to reduced credit demand and FX revaluation effects. Balance sheet growth has a floor of zero percent to prevent deleveraging (constraint binding in the adverse scenario). - RWAs projection: standardized and IRB portfolios differentiated. Standardized RWAs change due to balance sheet growth, new inflows of non-performing loans, new provisions, exchange rate movements, and conversion of some off-balance sheet items to on-balance sheet. IRB portfolios use through-the-cycle-PDs, downturn LGDs and EAD for each asset class/industry. - Interest income from non-performing loan is not accrued. - Capital actions: assume banks do not issue new shares or make repurchases during the stress test horizon. Dividends assumed to be paid out at 30 percent of current period net income after taxes (i.e., only if net income is positive) by banks in compliance with supervisory capital requirements. - Regulatory and market-based standards and parameters - National regulatory framework: Basel III regulatory minima on CET1 (4.5 percent) and include any requirements due to systemic buffers for three other systemically important institution (O-SII). - The team evaluates total banking capital adequacy ratio against the 8 percent level, Tier 1 capital ratio against the 6 percent benchmark and the leverage ratio against the 3 percent Basel III minimum requirement. - The same hurdle rate is used for baseline and adverse scenario. The hurdle rate for CET1, T1 and total capital adequacy do not include capital conservation and capital countercyclical buffers as well as pillar 2 requirement. - Banks that end the stress test horizon with a capital level or a leverage ratio below the relevant hurdle rates are considered to have failed the test. - Reporting format for results - Outputs include evolution of capital ratios for the system and groups of retail banks and large international banks. - Outputs include impact drivers (profit components, losses due to realization of different risk factors); capital shortfall as sum of individual shortfalls; capital shortfall in euros and in percent of nominal annual GDP; number of banks and corresponding percentage of assets below the regulatory minimum (or below the minimum leverage ratio). ### Banking Sector: Liquidity Stress Test — Top-down by IMF - Institutional perimeter - Exercise: Top-Down by FSAP team. - Institutions included: Seven banks subcategorized as SIs (same sample as in solvency stress test). - Market share: Total coverage is about 89.4 percent of the banking sector. - Data and baseline date: - Latest data: April 2023. - Source: supervisory data (LCR, NSFR, and ALMM Maturity Ladder template). - Scope of consolidation: banking activities of the consolidated banking group for banks having their headquarters in Belgium. Foreign subsidiaries assessed unconsolidated covering domestic activities only. - Channels of risk propagation / Methodology - Basel III LCR and cash-flow based liquidity stress test using maturity buckets by banks, incorporating both contractual and behavioral (where available) assumptions about combined interaction of funding and market liquidity and different levels of central bank support. - Liquidity test in EUR and foreign currencies (USD). - Risks and buffers - Risks: Funding liquidity and Market liquidity. - Buffers: Counterbalancing capacity, including liquidity obtained from markets and/or central bank facilities. Expected cash inflows included in cash-flow based and LCR-based analysis. - Tail shocks / Size of shock - Run-off rates calibrated to reflect system-wide deposit runs and dry-up of unsecured wholesale and retail funding, with additional run-off for non-resident deposits. Calibration follows historical events, recent international experience and IMF expert judgment. - Liquidity shocks simulated for 1–month for both LCR, and 5-days, 1-month, 3-months, and 1-year for the cash-flow based approach. - Haircuts of high-quality liquid assets (HQLA) calibrated against ECB haircuts, past Euro Area FSAPs, and market shock for investment securities and money market instruments in the solvency stress test. - Six cashflow analysis scenarios: - First three capture retail and wholesale stress translating into deposit outflows and lower inflows through losses from loans, with modest asset value effects affecting collateralized operations and liquid assets. - Fourth scenario features stress on wholesale market funding, translating into more aggressive haircut on assets. - Fifth scenario combines wholesale market stress with stress on wholesale deposits and loans. - Sixth scenario combines all previous five. - Sensitivity analysis of outflow rates on uninsured deposits to identify liquidity breaking point of the banking sector (retail deposits are main funding source). - Three cashflow stress scenarios: - Scenario A: deposits funding stress—household and corporates face strong liquidity strain and banks experience net outflows of deposits; counterbalancing haircuts increase mildly. - Scenario B: market liquidity stress—outflow of funds from wholesale borrowers and market price collapse of assets; CBC haircuts increase significantly. - Scenario C: combined stress in line with recent market turmoil / GFC. - Regulatory and market-based standards - Consistent with Basel III regulatory framework (LCR). - Liquidity shortfall assessed by bank. - Reporting format for results - Liquidity ratio or shortfall by groups of banks and aggregated (system wide). - Number of banks that still can meet or fail their obligations. ### Banking Sector: Solvency and Liquidity Interaction — Top-down by IMF - Institutional perimeter - Exercise: Top-Down by FSAP team. - Institutions included: Seven banks subcategorized as SIs (same sample as in solvency stress test). - Market share: Total coverage is about 89.4 percent of the banking sector. - Data and baseline date: - Latest data: April 2023. - Source: Top-down solvency and cashflow analysis output. - Scope of consolidation: banking activities of the consolidated banking group for banks having their headquarters in Belgium. Foreign subsidiaries assessed unconsolidated covering domestic activities only. - Channels of risk propagation / Methodology - Estimate the proportion of the HTM debt portfolio that will be sold to cover net outflows under the three cash-flow analysis scenarios. - Estimate unrealized losses due to revaluation of the price of these securities over the course of the scenario. - Assign an additional forced liquidation haircut and implement total forced liquidation losses to total profitability and capital depreciation over the course of the scenario. - Risks and buffers - Risks: Funding liquidity and Market risk. - Tail shocks / Size of shock - Total liquidity shocks over a period of 1-year for the cash-flow based approach. - Haircuts of HQLA calibrated against ECB haircuts, past Euro Area FSAPs, and market shock for investment securities and money market instruments in the solvency stress test. - HTM unrealized losses estimated using the historical increase of the risk rate in 2022 and the risk-free rate path of the macroeconomic scenarios. - Reporting format for results - Losses from HTM forced liquidation (system wide). - New profitability and capital adequacy ratios. ### System-wide: Interconnectedness Analysis — Top-down by IMF - Institutional perimeter and exercise - Exercise: Top-Down by FSAP team. - Institutions included: - Exercise A: Seven banks subcategorized as SIs (same sample as in solvency stress test, institution specific analysis). - Exercise B: Cross-border contagion using country-pair bilateral exposure across the world. - Exercise C: Aggregate domestic banking sector, insurance sector, and investment funds sector (sector-wide analysis). - Data: - Exercise A: Supervisory data: Bank balance sheet and supervisory statistics (COREP – Large exposures). - Exercise B: BIS consolidated banking statistics. - Exercise C: Leper cross-sectoral exposures. - Channels of risk propagation / Methodology - Balance-sheet model: Network model by Espinosa-Vega and Solé (2010). (Exercises A and B) - Satellite models on yields and share prices projections. (Exercise C) - Tail shocks / Size of shock - Exercise A: Default threshold—banks default if their capital falls below regulatory minimum. - Exercise B: Pure contagion—financial distress in foreign countries. - Exercise C: Market contagion through devaluation of assets. ### Insurance Sector: Solvency Analysis — Top-down by IMF (and Bottom-up by insurers) - Institutional perimeter - Exercise: Top-Down by FSAP team, Bottom-Up by insurance undertakings. - Institutions included: - 8 composite insurers (76% of total balance sheet assets). - 6 composite insurers (63% of total balance sheet assets) for BU exercise. - Data and baseline date: Regulatory reporting December 31, 2022. - Channels of risk propagation / Methodology - Investment assets: market value changes after price shocks affecting solvency position. - Insurance liabilities: impact on best estimate by changing discount rate of future cash flows. - Sensitivity analysis: effect on available capital and solvency position. - Tail shocks / Size of shock - Instantaneous shock: market shocks on equity and property prices front-loaded so maximum drawdown during project horizon realized immediately after reference date (end 2022). - Risk-free rate term structure adjusted to account for significant downward reversal of risk-free rates after 2023. Inflation shock extends CPI and wage inflation path to gradual return to two-percent inflation after the five-year model period. - A 30 percent lapse shock for non-mandatory insurance (term insurance, endowments, unit linked products, and disability) on insurers’ in-force life portfolio is assumed. - Specific shock magnitudes: - Risk-free interest rates (without volatility adjustment) 447 bps (1y EUR), 455 bps (10y EUR) - Sovereign bond spreads 90 bps domestic, 55 bps low spread EA countries, 120 bps high spread EA countries, 70 bps other advanced economies, 120 bps emerging and developing countries. - Stock prices -33.6 percent European Union, 33 percent other advanced economies, 32 percent emerging and developing economies. - Property prices 25 percent (commercial), 20 percent (residential) - Corporate bond spreads between 60 bps (AAA financials) and 355 bps (B and lower financials), and between 50 bps (AAA non-financials) and 325 bps (B and lower non-financials) - Mortgage default increase two percent domestic, three percent non-domestic - Reporting format for results - Impact on solvency ratios (with and without long-term guaranteed measures and transitionals). - Contribution of individual shocks to changes of eligible own funds. - Impact of reactive management actions (bottom-up only). - Dispersion measures of solvency ratios. ### Insurance Sector: Liquidity Analysis — Top-down by IMF - Institutional perimeter - Exercise: Top-down by FSAP team. - Institutions included: 8 composite insurers, 2 life insurers with significant IRS exposure (84% of total balance sheet assets). - Data and reference date: Regulatory reporting, December 31, 2022. *Source: Appendix II. Stress Test Matrix (STeM).* ### 8. Channels of ### 8. Channels of Risk Propagation ### Methodologies applied to liquidity and market channels - Derivatives revaluation: Revaluation of derivative position after interest rate shock. - Insurance/liquidity shocks: Mass lapse shock and shock to liquid assets. - Tail shock sensitivity: Parallel shift of the interest rate term structure, +25 bps, +50 bps, +100 bps. - Reporting metrics: - Variation margin as percent of cash holdings. - Variation margin as percent of cash holdings plus high-quality liquid assets. - Stressed liquidity ratios. ### Investment fund sector — top-down liquidity stress (Belgium) - Institutional perimeter: - Exercise: Top-Down by FSAP team. - Institutions included: Bond and mixed Investment funds. - Data sources: Lipper; NBB and FSMA supervisory data. - Portfolio reporting date: End of year 2022. - Channel and shock design: - Redemption shocks at various levels compared to highly liquid assets at fund-category level. - Redemption shocks calculated from historical redemptions using VaR and Expected Shortfall. - Methodologies with multiple thresholds; historical monthly time series. - Liquidation strategies: vertical vs. horizontal slicing. - Assumptions: First set calibrated on funds’ historical flows; additional shocks to asset values can be calibrated in line with the adverse scenario; price impact of asset sales assessed based on liquidation strategies. - Time horizon: Instantaneous shock. - Risk focus and outputs: - Main risk: Liquidity risk from severe but plausible redemption shock. - Buffer: Level of highly liquid assets. - Scenario analysis: Pure redemption shock based on historical distribution. - Market contagion: Diebold-Yilmaz methodology applied to fund indices to assess vulnerability to market distress. - Reporting format: - Number of funds with a redemption coverage ratio (ratio of highly liquid assets to redemptions) below one. - Liquidity shortfall amount for individual funds after redemptions. ### PDs, NPLs and solvency stress test inputs - Simulation approach: - All PDs and NPL ratios used in the solvency stress test estimated through time series modelling on the logistic transformation (logit) of variables. - Econometric models used to simulate 10,000 alternative paths under both the baseline and the adverse scenarios; the respective average paths used as projections. - NPL ratio model: - Historical NPL ratio estimated via single time series model with exogenous variables. - Sample: quarterly observations from Q1 1995 to Q4 2022 (112 observations). - Optimal specification: ARIMA(1,1,0). - Exogenous variables: one-period-lagged quarterly log-difference in property price index (PPI), one-period-lagged quarterly log-difference of CPI, and one- to three-period-lagged quarterly log-difference of real GDP. - Selected coefficient estimates (logit NPL model): - AR t-1: 0.252 (SE 0.098, tStat 2.583, pValue 0.011) - d LN PPI t-1: -0.713 (SE 0.374, tStat -1.910, pValue 0.059) - d LN CPI t-1: 1.694 (SE 0.803, tStat 2.109, pValue 0.038) - d Unemployment t-1: 0.029 (SE 0.012, tStat 2.354, pValue 0.021) - d LN GDP t-1: -0.264 (SE 0.361, tStat -0.730, pValue 0.467) - d LN GDP t-2: -0.287 (SE 0.381, tStat -0.755, pValue 0.452) - d LN GDP t-3: -0.604 (SE 0.345, tStat -1.753, pValue 0.083) - Corporate and financial EDFs: - Corporate EDF: - Historical Moody’s average corporate EDF of 11 countries estimated via panel ARIMA(1,1,0) with exogenous variables. - Sample: quarterly observations Q2 2013 to Q4 2022; panel unbalanced; 417 observations. - Exogenous variables: country-specific quarterly change in sovereign spreads and quarterly log-difference real GDP. - Selected coefficient estimates: - AR t-1: 0.2600 (SE 0.0471, tStat 5.5155, pValue 0.0000) - d LN GDP t: -0.4135 (SE 0.1717, tStat -2.4088, pValue 0.0164) - d SovSpread t: 0.0517 (SE 0.0157, tStat 3.2975, pValue 0.0011) - Financial institutions EDF: - Panel ARIMA(1,1,0) with exogenous variable one-period-lagged quarterly change of output gap. - Sample: quarterly observations Q2 2013 to Q1 2023; panel unbalanced; 429 observations. - Selected coefficient estimates: - AR t-1: 0.1329 (SE 0.0480, tStat 2.7666, pValue 0.0059) - d OutputGap t-1: -0.0102 (SE 0.0055, tStat -1.8376, pValue 0.0668) ### Interest rate regressions (loan and deposit rates) - Loans: - Historical interest rates for new/variable residential mortgage loans, unsecured consumer loans and corporate loans are cointegrated and modeled jointly via a vector error correction model (VECM) with quarterly change of EURIBOR as exogenous. - Sample: quarterly observations from Q1 2006 to Q4 2022 (70 observations). - Optimal order: zero (per BIC). - Selected coefficients (VECM impacts and d EURIBOR t effects): - Constant (Retail Mortgage): 0.1221 (SE 0.0915, pValue 0.1820) - Constant (Consumer): 0.8333 (SE 0.1782, pValue 0.0000) - Adjustment (Consumer, Retail Mortgage): 0.2269 (SE 0.0450, pValue 0.0000) - Impact (Consumer, Retail Mortgage): 0.5280 (SE 0.1047, pValue 0.0000) - Impact (Consumer, Consumer): -0.4628 (SE 0.0918, pValue 0.0000) - d EURIBOR t (Retail Mortgage): 0.2094 (SE 0.0576, pValue 0.0003) - d EURIBOR t (Consumer): 0.0629 (SE 0.1122, pValue 0.5749) - d EURIBOR t (NFC Loans): 0.4314 (SE 0.0905, pValue 0.0000) - Deposits: - Four deposit-rate models (term deposits, retail overnight, corporate overnight, savings), single-equation time series with d EURIBOR and risk-free rate as exogenous, and AR components. - Sample: quarterly Q1 2006 to Q4 2022 (70 observations). - Selected estimates: - AR t-1 estimates: Term -0.3036 (SE 0.0981, pValue 0.0030); Retail overnight -0.0751 (SE 0.0924, pValue 0.4196); Savings 0.2133 (SE 0.0985, pValue 0.0342). - AR t-2: Retail overnight 0.3612 (SE 0.0840, pValue 0.0001). - d EURIBOR t estimates: Term 0.9609 (SE 0.1092, pValue 0.0000); Retail overnight 0.1113 (SE 0.0161, pValue 0.0000); Corporate overnight 0.2782 (SE 0.0386, pValue 0.0000); Savings 0.1904 (SE 0.0350, pValue 0.0000). - d RiskFree t: Retail overnight 0.0590 (SE 0.0334, pValue 0.0817). ### Combining PD estimates - Combined PD formulation: - PD_t^c = w_a PD_t^a + w_b PD_t^b subject to w_a + w_b = 1. - Combination weights estimated via linear programming calibrated by in-sample forecasting accuracy; observed and predicted PDs normalized before weight calibration. - Linear program minimizes sum of positive and negative deviations across T observations with non-negativity constraints. - Estimated combination weights: - 29.5 percent for the NPL base projection. - 70.5 percent for the EDF based projection. ### Liquidity Coverage Ratio (LCR) stress scenarios - LCR definition and approach: - LCR computed as ratio between weighted average of high-quality liquid assets (HQLA) and weighted average of inflows and outflows over 30 calendar days. - Baseline LCR uses Basel III prescribed inflow/outflow/asset coefficients (“Basel”). - Stress simulated by proportionally changing coefficients: lowering inflows, increasing outflows and haircuts on liquid assets. - Six simulated scenarios calibrated to incorporate mild and severe stress on asset values, inflows, and outflows. - Under “aggressive” scenarios stressed Basel coefficients are roughly doubled/halved versus Basel coefficients. - Assumption: no change to asset composition or flows when stressing weights. ### Cashflow and insurance stress test calibrations - Cashflow scenarios: calibrated cashflow tables and scenarios used for liquidity stress calibrations (Appendix VII). - Insurance stress test market and credit risk calibration summarized in tables (Appendix VIII); country grouping note: Low spread, high-rated EA countries include Austria, Finland, France, Germany, Ireland and Luxembourg and the Netherlands. ### Cross-sectoral interconnectedness and iteration algorithm - Regression framework: - Four equations estimated (two for banks: bond yields and share prices; two for insurers: bond yields and share prices). - Bank sample: yields on bank-issued debt (maturing between end 2025 and end 2026) for six banks; share-price regressions for three banks. Sample period: 2010Q1–2022Q4. Data sources: Bloomberg for yields/prices; Fitch for balance-sheet regressors. - Insurer sample: three insurers (AG Insurance, Baloise, Ethias); frequency annual; sample period 2010–2022. - Key regressors: - Capitalization proxied by equity-to-assets ratio. - Credit quality proxied by NPLs/total loans, loan loss provisions/total assets, log loan loss reserves/loans. - Liquidity proxied by liquid assets/total assets. - Controls: 10y Bund (risk-free), bid-ask spread, interbank spread, stock volatility index, change in gross written premia. - Selected regression coefficient highlights (Banks and Insurers; robust SEs in brackets): - Share Prices (Banks) lag L1: -0.440*** [0.145] - D(10y Bund): 0.674** [0.264] - D(Equity/Assets) (Yields): -0.630** [0.296] - D(NPLs/Loans): 0.871** [0.415] - D(LOG(LLR/Loans)): -2.650* [1.469] - D(Liquid Assets/Assets): -0.057 [0.043] (Banks) and 2.639** [1.243] (Insurers) - D(VStoxx50): 0.081*** [0.028] (Yields) and -0.507*** [0.156] (Share Prices) - Observations: Banks equations: 1238 and 02617? (table shows concatenated numbers—observations listed as 1238 02617); Insurers observations: presented in table. - R-squared reported examples: 0.169, 0.237, 0.502, 0.532. - Significance markers: *** p<0.01, ** p<0.05, * p<0.1. - Stress transmission mechanism: - Solvency stress-test outputs (changes in capitalization, NPLs, LLPs, liquidity) are input into regressions to predict changes in bond yields and share prices over the stress horizon (banks: four-year scenario; insurers: instantaneous shock). - Investment funds’ price changes assumed proportional to changes in other asset prices weighted by funds’ exposures (Who-to-whom matrix, end 2022). - Losses from asset-price changes applied to banks’ and insurers’ equity using average modified duration for bonds; exposures from Who-to-whom matrix. - Equity and assets updated; new equity-to-assets ratio used to compute second-round effects on yields and share prices via the same regressions. - Iteration: algorithm iterated until convergence; converged after about ten iterations as losses and asset prices stabilized. ### Investment funds sample definition - Funds included: collective investment undertakings in Economic Function 1 susceptible to runs — fixed income funds, mixed funds, funds-of-funds (FoFs), and money market funds. - Rationale: These vehicles may engage in maturity/liquidity transformation or employ leverage and could become susceptible to liquidity pressures because of heightened investor redemption requests. *Source: IMF Staff* ### 2. Overall, the analysis considered 435 funds with different investment policies and ### 1belea2023011 - 2. Overall, the analysis considered 435 funds with different investment policies and ### Data - Sample: 435 funds with different investment policies and objectives, including bond funds, mixed-FoFs, non-public alternative investment funds (sold to professional investors), and a residual category (structured funds, pension saving funds that are collective investment undertakings under the Belgian law, and money market funds). - Structured funds: 113 funds amounting to 3 EUR bn. - Residual category of other funds: NAV of 31.6 EUR bn, including 7 EUR bn in money market funds. - Mapping and sources: - Investment funds mapped in Lipper using ISINs and LEIs. - Monthly data on flows, net asset value, returns, investment objectives and geographical focus retrieved over 2008–2021. - Overall, 390 investment funds included in the analysis are retrieved in Lipper. - Ancillary information on asset characteristics, sectors, credit quality, maturities, yields, and durations from Refinitiv EIKON and Bloomberg. - Metrics on liquidity and trading derived from Eikon and Bloomberg. - Total volumes of domestic sovereigns taken from MTS. - Computation of net flows: - Net flows in percent of NAV computed for each fund using FNFFS_t / NAV_{t−1} (equation (29) as presented). - Monthly flows aggregated at portfolio level across different share classes. - Net flows whose absolute value is above 50% were excluded as likely related to mergers and liquidations of share classes or reporting errors. - Portfolio composition: - Supervisory information on fund characteristics and asset-level portfolio composition at the end of 2022 used, including asset categories and ISINs, price and market value, portfolio weights, amount of cash held. - Look-through approach applied to other collective investment undertakings; underlying funds’ portfolio composition obtained from Lipper. ### Calibration of the redemption shock - Objective: assess liquidity risk for open-ended investment funds exposed to fixed income instruments by calibrating a plausible but severe instantaneous redemption shock and comparing it to highly liquid assets. - Methodology references: follows Bouveret and Yu (2021, IMF) and analogous FSAP approaches (IMF, 2015b; 2017, 2018, 2021, 2022). - VaR-based calibration: - VaR at level α: FVaR(α) = F^{-1}(α) (equation (30) as presented). - Limitations noted: extreme shocks below the VaR not considered; parametric VaR subject to model risk. - Expected Shortfall (ES) calibration: - ES defined as average net flows below the VaR: NES(α) = E(Z | Z < FVaR(α)) (equation (31) as presented). - Treatment for funds not in Lipper: shock equals average of shocks for comparable funds by investment policy, geographical focus and ownership structure. - Shock nature: instantaneous (no persistence over several periods). - Assumptions and calibration focus: - Homogeneity assumption: each fund within same investment type and same geographical focus faces same redemption shock; calibration based on the 3 percent ES. Robustness checks: ES at 1 percent and 5 percent, and VaR at worst 1 percent, 3 percent, and 5 percent net flows observed. - Heterogeneity assumption: shock calibrated separately for each fund based only on its own historical data; main shock based on 3 percent ES; robustness checks at 1 percent and 5 percent and using percentiles. - Limitations of heterogeneity: not comparable across funds and may be uninformative if a fund never experienced large outflows. - Each fund is subject to 12 different redemption shocks. Main focus: shocks calibrated at the 3 percent level. - Appendix X. Table 1: Calibration of Redemption Shocks (selected exact figures preserved) - Level: 1% - Bond funds: Homogeneous ES 20%, VaR 18%; Global focus ES 19%, VaR 15%; Heterogeneous ES 17%, VaR 13%. - Mixed (FoFs): Homogeneous ES 9%, VaR 2%; Global focus ES 15%, VaR 9%; Heterogeneous ES 6%, VaR 5%. - Non-public alternative funds: Homogeneous ES 24%, VaR 24%; Global focus ES 21%, VaR 17%; Heterogeneous ES 17%, VaR 16%. - Other: ranges reported (e.g., Homogeneous ES 1%-40%, VaR 1%-39%; etc.). - Level: 3% (main focus) - Bond funds: Homogeneous ES 15%, VaR 10%; Global focus ES 14%, VaR 9%; Heterogeneous ES 12%, VaR 7%. - Mixed (FoFs): Homogeneous ES 4%, VaR 2%; Global focus ES 9%, VaR 5%; Heterogeneous ES 5%, VaR 3%. - Non-public alternative funds: Homogeneous ES 18%, VaR 11%; Global focus ES 14%, VaR 8%; Heterogeneous ES 15%, VaR 8%. - Other: ranges reported (e.g., Homogeneous ES 1%-30%, VaR <1%-29%; etc.). - Level: 5% - Bond funds: Homogeneous ES 12%, VaR 7%; Global focus ES 12%, VaR 7%; Heterogeneous ES 10%, VaR 6%. - Mixed (FoFs): Homogeneous ES 3%, VaR 2%; Global focus ES 7%, VaR 4%; Heterogeneous ES 4%, VaR 2%. - Non-public alternative funds: Homogeneous ES 14%, VaR 7%; Global focus ES 11%, VaR 3%; Heterogeneous ES 11%, VaR 2%. - Other: ranges reported (e.g., Homogeneous ES 1%-33%, VaR <1%-25%; etc.). - Note: Redemption shocks defined as net outflow in % of NAV. Average flow by fund category under the heterogeneity assumption. Monthly flows from January 2008 to December 2022. Source: Lipper, IMF staff. ### Investment funds’ resilience: the liquidity bucket approach - Measurement: compare redemptions to level of highly liquid assets using liquidity weights from the Liquidity Coverage Ratio (Basel Committee) to allow comparability. - Redemption Coverage Ratio (RCR) defined (equation (32) as presented). Interpretation: - When RCR < 1, the fund does not have enough highly liquid assets to cover redemptions without selling less liquid assets. - Highly liquid assets for fund i: - HLA_i = Σ_{r=1}^{L} ω_{i,r} × s_{i,r} (equation (33) as presented), where ω_{i,r} are liquidity weights assigned to each security s_{i,r}. - Liquidity shortfall when RCR < 1: - Shortfall = Redemption shock − HLA (expression in text corresponding to equation (34) as presented). - Appendix X. Table 2: Liquidity Weights (exact weights preserved) - Cash: 100% - Equities: 50% - Sovereign bonds: AAA−AA 100%; A 85%; BBB 50%; Below BBB 0% - Corporate bonds: AAA−AA 85%; A 50%; BBB 50%; Below BBB 0% - Covered bonds and Securitized: AAA−AA 85%; A 0%; BBB 0%; Below BBB 0% ### Liquidation strategies and price impact of funds sales - Liquidation strategies considered: - Vertical slicing (pro rata): sell each asset class in proportion to their portfolio weight (preserves portfolio composition). - Waterfall: sell most liquid assets first (may mitigate price impact but leaves remaining investors with a less liquid portfolio). - Order of liquidation based on HQLA liquidity weights; unrated instruments liquidated after assets with positive liquidity weights are exhausted. - Observed behavior in COVID-19 turmoil: funds increased cash positions while decreasing portfolio share in sovereign and other high-quality assets. - Market depth and price impact estimation: - Market depth MP(τ) = HPF / σ × √τ (equation (35) as presented), where τ is time horizon to sell assets, HPF average daily trading volumes, σ asset volatility. - Given instantaneous shock, τ = 1 day. - Price impact calculated for domestic sovereigns: volatility estimated by filtering ICE BofA All Maturity Belgium Government Index market index through a GARCH (1,1) and taking maximum historical volatility. - Average monthly traded volumes at end of 2022 taken from MTS. - Price impact metric: NI(τ) = Sales / MP(τ) (equation (36) as presented). ### Investment funds stress test results - Results presented for 12 redemption shocks (ES and VaR at 1%, 3%, and 5%) under homogeneous and heterogeneous assumptions. - Appendix X. Table 3: Investment Fund Stress Tests Results: Resilience (share of investment funds with RCR < 1; figures preserved) - Level: 1% - Bond funds: Homogeneous ES 4.3% (VaR not reported in table cell as dash). - Mixed (FoFs): Homogeneous ES 3.3%, VaR 0.5%; Heterogeneous ES 1.4%, VaR 0.5%. - Non-public alternative funds: Homogeneous ES 26.3%, VaR 21.1%; Heterogeneous ES 21.1%, VaR 21.1%. - Other: dashes (no values reported). - Level: 3% - Bond funds: dashes (no values reported). - Mixed (FoFs): Homogeneous ES >1% (VaR dash); Heterogeneous ES >1% (VaR dash). - Non-public alternative funds: Homogeneous ES 13.2%, VaR 5.3%; Heterogeneous ES 15.8%, VaR 15.8%. - Other: dashes. - Level: 5% - Bond funds: dashes. - Mixed (FoFs): dashes. - Non-public alternative funds: Homogeneous ES 5.3%, VaR 5.3%; Heterogeneous ES 13.2%, VaR 5.3%. - Other: dashes. - Note: Share of investment funds with RCR < 1. Source: IMF staff. ### Volatility spillover analysis - Objective: analyze transmission of volatility shocks between fund categories and underlying markets using weekly returns for equity, bond and mixed funds. - Methodology: Diebold and Yilmaz’s (2009, 2012) approach; VAR estimated using a lasso-estimator. - Definitions and model: - VAR specification: X_t = Φ(N) x_t + ε_t (equation (37) as presented). - Generalized forecast-error variance decomposition (FEVD) used with Pesaran and Shin’s (1998) methodology; order invariant. - Connectedness measures s̃_ij(H) computed from FEVD (equation (38) as presented). - Spillover from i to j: percent of j’s total inward spillovers attributable to i (conceptual definition preserved). - Data: - Weekly returns retrieved over June 2014–December 2021. - Volatility for each fund or market index estimated via GARCH (1,1). - Market indices considered: - Fixed income markets: - ICE BofA All Maturity Belgium Government Index - ICE BofA Euro Corporate Index - ICE BofA Global Corporate (Excluding Euro) Index - Stock markets: - BEL20 - EUROSTOXX600 - SP500 *Source: IMF staff (extracted from 1belea2023011 - 2. Overall, the analysis considered 435 funds with different investment policies and).* --- _Source: https://www.imf.org/-/media/files/publications/cr/2023/english/1belea2023011.pdf_