## 1cheea2025011-source-pdf ## Source details **Canonical URL:** [1cheea2025011-source-pdf](https://www.imf.org/-/media/files/publications/cr/2025/english/1cheea2025011-source-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/cr/2025/english/1cheea2025011-source-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/cr/2025/english/1cheea2025011-source-pdf.pdf.json) --- ### Executive summary — shocks, environment, and vulnerabilities - Recent shocks and crisis management: - Turbulent period since 2019 FSAP: COVID 19 pandemic, geopolitical conflicts, and the collapse of Credit Suisse (CS) in 2023. - Government-assisted UBS–CS merger entailed significant contingent fiscal liabilities, undermined TBTF credibility, and revealed gaps in supervision, resolution, and crisis management. - Structural vulnerabilities: - Rising concentration and competitive distortion risk. - Periods of strong Swiss Franc appreciation weighing on exporters and complicating monetary policy. - High household debt and large banks’ mortgage credit exposure. - High and rising house prices and housing affordability strains. - Weak banking profitability driven by high operating costs and low interest rates. - System size and concentration (end-2023 / end-2024 snapshots): - Financial system size ≈ 1,000 percent relative to Swiss nominal GDP (split: about 600 percent for NBFIs and 400 percent for the banking system). - Four largest banks (SIBs: one GSIB and three DSIBs) represent 264 percent of GDP and 65 percent of banking system assets. - Number of banks fell from 275 in 2014 to 213 in 2024 (22 percent decline). - Banking system assets shrank from 500 percent relative to GDP in 2021 to 400 percent in 2024. - SIX Stock Exchange free float market capitalization ≈ CHF 1.8 trillion (225 percent of GDP, at end-2024). - Nonfinancial private sector debt: 267 percent relative to nominal GDP in 2024. - Pension funds: assets around 140 percent relative to GDP; second Pillar sector size CHF 1.2 trillion or 152 percent of GDP by end-2023. ### House prices, mortgage structures, and risks - Valuation and model estimates: - SNB estimates house price overvaluation ranges between 15–40 percent. - Price-to-income and price-to-rent ratios indicate house prices rest by about 30 percent above historical averages. - Model point estimates between 15-25 percent; upper confidence bands up to 35 percent. - Conservative estimate between point and upper bound used for stress scenarios. - Structural drivers supporting mortgage debt: - Tax incentives, interaction with pension saving schemes, lenient amortization requirements. - Five supporting factors preserved verbatim: 1. At origination, mandatory pension savings (Pillar 2) and voluntary pension savings (Pillar 3) can be pledged for mortgage borrowing, allowing higher leverage. 2. At origination, 10 percent of the lending value of the property must be covered by the borrower‘s own funds; the remaining part of the down payment can be covered by depleting Pillar 2 or 3 savings (for Pillar 2, up to 100 percent of the savings until age 50; the higher of up to half of it or the amount at 50 after age 50). 3. The option to have up to two thirds of total lending value not amortizing. 4. Pillar 3 savings can be used for “indirect amortization,” i.e., borrowers save in a Pillar 3 scheme to repay the mortgage at maturity while retaining tax benefits from interest deductibility and generating interest income and capital gains from Pillar 3 investments. 5. Interest on mortgage debt is tax deductible against income. - Non-amortizing mortgages share for lowest income buckets reaches up to 70 percent. - Mortgage lending grew to represent 37 percent of total assets at end-2024 (30 percent at end-2019; mortgages constitute 66 percent of the loan book). - Variable rate mortgage shares have risen to 20 percent. - Risks and concentrations: - CRE lending reaches 25-30 percent of loan books for some cantonal and regional banks. - Real estate-related investments account for 31 percent of life insurers’ investment portfolios; pension funds hold 23 percent in real estate-related investments. - UBS is the world’s largest G-SIB relative to domestic GDP after the CS takeover. ### Bank solvency stress testing — design, results, caveats - Sample and scope: - 92 (112) banks at consolidated (solo) level; covers 93 percent of Swiss banking system assets at end-2024. - Clusters: four SIBs (one aggregate Raiffeisen), 23 cantonal banks, 12 asset and wealth management banks, 42 regional banks, 11 other banks. - Model architecture and key inputs: - Solvency model mixes structural and econometric elements; seven credit portfolios with PD and LGD models. - Household mortgage PD/LGD from micro-macro simulation model IDHBS+ (SILC 2022 and HBS). - Market risk: HFT and AFS revalued using modified duration; HTM not marked-to-market in main run (sensitivity revalued HTM). - PD and LGD trajectories: - Household mortgage LGDs rise from about 10 percent to 30 percent under the demand shock scenario. - CRE LGDs rise from 15 percent to more than 50 percent at peak. - Scenario framing: - Horizon: three-year period 2025–27. - Two adverse scenarios: Adverse (D) demand shock (disinflationary, policy rates reduce); Adverse (S) supply shock (inflationary, policy rates increase). - Business-cycle severity multipliers (std dev multiples, year 1 and cumulative 2 years): - Demand shock scenario: 2.4x and 2.3x. - Supply shock scenario: 1.7x and 1.9x. - Aggregate results: - Aggregate CET1 for 92 banks moves from 17 percent at end-2024 to 10.8 percent at the trough. - Capital ratio of the aggregate sample drops by 6.2 p.p. to the low point in Year 2. - Cluster drops (start and drop): - Cantonal banks’ solvency least affected: -4.3 pp. - Asset managers: starting CET1 20 percent and drop -13.6 pp. - SIBs aggregate CET1 ratio drops -6.3 pp; all SIBs remain above hurdle rates. - Drivers of decline (supply shock, up to third year end-2027, 3-year cumulative): - Loan losses: -4 pp. - Fee and commission income: -2.2 pp. - Trading income: -1.7 pp. - Net interest income: -1.4 pp. - Shortfalls and systemic significance: - Five of 92 banks fall below CET1 hurdle rate by the third year of the supply shock. - Six banks fall short on total capital ratios. - Combined assets of these banks represent 4.6 percent of banking system assets at end-2024. - Combined CET1 capital shortfall = 0.4 percent of GDP under supply shock Year 3 (0.04 percent of GDP under demand shock). - Caveats and uncertainty: - High uncertainty for asset and wealth management banks: starting risk parameters for Lombard loans, extent of hedging, and dynamics of fee & commission income not fully observed. - Recommendation: Enhance data coverage and include asset and wealth management banks in regular stress tests. ### Bank liquidity stress testing — design, results, and feedbacks - Model design: - Cash flow-based liquidity model using detailed maturity profiles, runoff/drawdown assumptions, and counterbalancing capacity (cash and reserves, interbank deposits, reverse repos, bond and equity holdings). - Run-off assumptions for liabilities maturing within 30 days (cumulative): - Household sight deposits: 10 percent. - Household term deposits: 30 percent. - Deposits from NFCs, banks, NBFIs, bonds: between 30 and 40 percent. - Secured funding: 15-25 percent. - Open committed credit lines drawn by 20 percent within first four weeks. - Results (sample: 92 banks): - Baseline liquidity simulation: 10 banks estimated to experience liquidity shortfall before 30 days; all 10 are regional and other banks. - Liquidity shortfall of failing banks accumulates to 0.15 percent of initial liquid assets of all 92 banks. - Severe valuation stress (six-month horizon): up to 19 banks face shortfalls, mostly small regional banks. - System-level metrics: - All banks satisfy LCR requirements; cantonal, regional, and other banks have lower LCRs. - NSFRs comfortable at median across clusters; higher for asset/wealth managers. - Unencumbered liquid assets as percent of total liabilities lowest for regional and other banks. - HTM sales feedback: - Feedback from HTM bond sales to solvency limited generally. - Under supply shock extension (HTM loses value at outset): at 6-month horizon, 10 banks face feedback less negative than -0.5 p.p.; five banks face feedback more negative than -1.5 p.p. - Affected banks include regional banks and asset/wealth managers. ### Insurance sector stress testing — methodology and results - Scope and baseline: - Six Swiss insurance groups covering about two thirds of Swiss life and non-life market and 70 percent of reinsurance sector; one primarily domestic, five IAIGs. - Investments ≈ 60 percent of balance sheet; unit-linked adds another 25 percent. - More than 90 percent of bonds are investment grade; around 56 percent are AA or higher. - Stress approach: - BU (internal models) and TD approaches; instantaneous shock at start of Year 1 with three-year projections. - Market shocks front-loaded; reactive management allowed in BU not in TD. - Results: - BU: median SST ratio drops from 255 percent before stress to 219 percent after stress; no insurer fell below regulatory minimum in BU. - TD aggregate: SST ratio fell by more than 40 percentage points (size-weighted), from baseline 236 percent; balance sheet shrank by CHF 180 billion post-stress; equity (excess of assets over liabilities) loss nearly CHF 60 billion. - Conclusion: industry can withstand severe price shocks though solvency ratios fall significantly. - Interest rate and liquidity dynamics: - Average negative duration gap ≈ two years; net valuation effect of rising rates is positive for risk-bearing capital. - Insurers exposed to margin calls and surrenders; mitigants include repo market access, SNB standing facilities, contractual surrender restrictions, and low asset encumbrance. - Surrender payout timing and deductions limit immediate liquidity pressure; best-estimate liquid assets exceed total surrender values by a considerable margin. - Cyber sensitivity: - Cyber premium volume around CHF 450 million in 2020, growing >130 percent over three years; remains <1 percent of total business. - Ransomware attack scenario: average fall in SST ratio of five percentage points. - CSP outage scenario: average fall in SST ratio of three percentage points. ### Pension funds — structure, risks, and findings - Size and concentration: - Assets around 140 percent of GDP; Second Pillar sector CHF 1.2 trillion (152 percent of GDP) by end-2023. - More than 1,300 occupational pension institutions; top 14 percent largest entities represent 82 percent of active members and total pension assets. - Defined-contribution regimes represent almost 90 percent of pension liabilities. - Pension funds’ domestic real estate holdings CHF 217bn (insurance sector CHF 67bn). - Funding and solvency: - Capital weighted ratio for pension funds without state guarantee = 115.6 percent in H1 2024. - Number of funds with coverage ratio below 100 has fallen to less than one percent. - Coverage ratios and fluctuation reserves: target fluctuation reserves on average 2 percent lower for those with state guarantee; technical interest rates higher for state-guaranteed funds. - Market dynamics and risks: - Semi-autonomous solutions gaining importance; autonomous funds still dominant (35 percent of funds but ~70 percent of liabilities). - DC regimes dominate (~90 percent of liabilities); hybrid cash balance plans with minimum guaranteed interest rate and fixed conversion features. - Lump sums and withdrawals: minimum 25 percent can be taken as lump sum; complete withdrawals >40 percent in 2023; partial withdrawals another 20 percent. - Conversion rate: statutory conversion rate unchanged since 2014; conversion rate of 6.8 percent presents sustainability challenges. - September 2024 Pillar 2 reform referendum rejected by two-thirds majority; immediate sector impact low but may accelerate consolidation. ### Climate physical risk — methods and impacts - Flood risk analysis focus: - RCP 8.5 scenario; worst-case flood events from Munz et al. (2023); event and future-condition hazard indicators used. - Data layers: residential building locations, SILC and HBS microdata, banks’ canton-level mortgage distributions. - Channels to bank credit risk: - LGD channel: acute physical damage reduces property values; LGDs increase materially (CRE and mortgage LGDs modeled linked to price impacts). - PD channel: increase in insurance expenses mapped into household consumption and hence PD increases. - Key results: - Extreme flood scenario: LGD channel affects bank capital more than PD channel. - Cantonal and regional banks most affected due to greater mortgage exposures and regional concentration. - If insurance coverage of losses prevails, LGD impacts are nullified. - Insurance-transfer mechanisms: - Swiss building insurance systems (KGVs in 19 cantons; private insurers in GUSTAVO cantons) operate under double solidarity (among insureds and among insurers) with inter-cantonal funds and compensation funds; prevention and response measures included. - Recommendations: Collect more data from financial institutions to quantify climate-related risks and collaborate with FOEN, building insurers, and other authorities. ### Interconnectedness, contagion analysis, and holistic vulnerability assessment - Methodology: - Domestic bilateral interbank exposures and Espinosa‑Vega & Solé (2010) default simulations; BIS locational cross-border exposures for international contagion. - Key findings: - Banks highly exposed to OFIs; OFIs may present contagion risk. - Switzerland cross-border claims USD 706 billion as of 2024Q3; liabilities USD 633 billion. - Hypothetical default of SIBs causes most notable domestic spillovers; cantonal banks appear more vulnerable than other clusters. - Switzerland ranks 18th most vulnerable to cross-border spillovers among 29 jurisdictions; ranks 10th on impact metric (contagion from Switzerland is low). - Holistic vulnerability ranking: - Combines 20 metrics (25 percent initial conditions; 75 percent forward-looking). - Forward-looking analysis makes composite rankings resemble forward-looking ranking. - Residual cluster comparably weakest on initial conditions; asset managers and SIBs appear stronger forward-looking. - One bank identified as both impactful and vulnerable in interconnectedness metrics. ### Data, model development, and supervisory recommendations (selected) - Data collection and monitoring recommendations (verbatim where possible for main list items and timings): - 1 Keep monitoring housing and mortgage market developments, including house price valuation, affordability strains and interest rate risk for households, and consider related macroprudential policies to contain systemic risk. - Timing: ST/C; Responsible: FINMA, SNB - 2 Monitor and collect more data pertaining to commercial real estate business, real estate (RE) firms’ borrowing from banks and capital markets, to assess their relevance and contribution to systemic risk and vulnerabilities vigorously and continuously. Also consider developing transaction-based price indices for commercial RE. - Timing: ST/C; Responsible: FINMA, SNB - 3 Compile bilateral exposure data involving all financial institution types, beyond banks, for authorities’ network and contagion analysis to achieve a fuller system coverage. - Timing: MT; Responsible: FINMA, SNB - 4 Collect more data on a regular basis to enhance the risk analysis and modeling, including related to mortgages for banks other than SIBs, banks’ trading and investment portfolios, banks’ foreign exposures in general, and Lombard loan exposures specifically. - Timing: MT; Responsible: FINMA, SNB - 5 For insurers, reduce dependence on ad-hoc data requests and establish more frequent regulatory reporting. - Timing: MT; Responsible: FINMA - 6 Develop a dataset and regularly collect data for pension funds to enable market-wide horizontal and systemic risk analyses, particularly regarding asset exposures. - Timing: MT; Responsible: OPSC - 7 Collect more data from financial institutions to quantify climate-related risks. The authorities should collaborate with the Federal Office of Environment, building insurers, other relevant authorities, and private sector entities to monitor such risks. - Timing: MT; Responsible: SNB, FINMA - 8 Develop a bank liquidity stress test model, to complement the monitoring based on data and the analysis of regulatory metrics as so far prioritized. - Timing: MT; Responsible: SNB, FINMA - 9 Keep developing micro data-based models for nonfinancial corporates and households and connect them to the bank solvency stress test model. - Timing: MT; Responsible: SNB - 10 Develop additional model structures pertaining to asset and wealth management banks, to include them in the stress test models and thereby support the analysis of their vulnerabilities and systemic relevance. - Timing: MT; Responsible: SNB, FINMA - 11 Benchmark and assess the banks’ provisioning practices for performing exposures. - Timing: MT; Responsible: FINMA, SNB - 12 Conduct liquidity-focused top-down stress tests and scenario analysis for insurers. - Timing: MT; Responsible: FINMA - 13 Ensure comparability of the calculation of pension funds’ funding position. Provide cantonal supervisors with the necessary authority and instruments to intervene. - Timing: MT; Responsible: OPSC, cantonal supervisors, Federal Council - 14 Consider providing more information about the macro-financial scenarios and bank solvency stress test results in SNB’s financial stability report. - Timing: MT; Responsible: SNB - 15 FINMA and the SNB should collaborate regarding the reciprocal benchmarking of top-down and bottom-up stress test results for banks. - Timing: MT; Responsible: FINMA, SNB - 16 Establish a common platform to assess and discuss market/sectoral developments and potential risks to financial stability, the findings of which should feed into the Steering Committee/Standing Committee for financial stability discussions. - Timing: MT; Responsible: OPSC, Federal Council, cantonal supervisors, FINMA, SNB - Additional supervisory recommendations (selected): - FINMA should keep examining banks’ valuation models for real estate collateral at origination and along loan lifetimes. - Consider abolishing the tax deductibility of mortgage interest payments (reiterated from 2019 FSAP). - SNB should publish more details in its FSR related to macro-financial scenarios and stress test results. - FINMA to implement top-down liquidity stress testing for insurers and collect detailed derivatives and surrender/lapse data with higher reporting frequency. - Authorities to reduce reliance on external regulatory auditors for pension funds and grant supervisors the authority for on-site inspections on a risk-based approach. ### Scenario design summary (baseline and adverse) - Horizon: 2025–2027 (three-year). - Scenarios: baseline (aligned with IMF WEO, IMF-internal one as of February 2025), Adverse (D) demand shock (disinflationary, policy rates reduce), Adverse (S) supply shock (inflationary, policy rates increase). - Market shock magnitudes (insurance tail specification examples): - Equity domestic: -45.0 percent. - Domestic Residential property: -25 percent; Domestic Commercial: -30 percent. - Mortgage and loan default rates: 150 (basis points increase). - CHF appreciation vs EUR, USD, JPY: 15 percent; GBP: 20 percent. - Risk-free interest rate shocks (bps): short-term CHF 107; long-term CHF 337. - Corporate bond spread examples (bps): AAA 60-75; A 240-275; BBB 290-310; BB 440-465; B and lower up to 560-570 depending on instrument and jurisdiction. - Business-cycle severity multipliers (std dev multiples): - Demand shock: 2.4x and 2.3x. - Supply shock: 1.7x and 1.9x. ### Model and data infrastructure notes (appendices) - Household sector model (IDHBS+): - Data: SILC (2022) 8,900 households and 19,229 household members; HBS 2015-2017 9,955 households. - Key parameters: Replacement rate 72 percent; Maximum unemployment benefit per month CHF 7,970; Variable rate share of total household debt 20 percent; PD anchor for mortgage debt 0.5 percent; LGD anchor for mortgage debt 10 percent; Cure rate 20 percent. - Corporate PD models: - Bayesian Model Averaging (BMA) used; segments NFC (307 Swiss firms), CRE (30 firms), Banks (32 firms), NBFIs (63 firms). - Historical fit examples: NFC median PD including stock prices R-square = 0.98; CRE R-square = 0.59; Banks R-square = 0.85. - Stress test perimeter and cut-off: 92 banks (consolidated) representing 93 percent of assets; regulatory cut-off date end-December 2024. - Liquidity run-off assumptions and haircuts aligned with LCR regulatory assumptions and CS run-off experience. - Climate flood scenarios: nine flood events from Munz et al. (2023); RCP 8.5 and 2-degree GWL scenarios used for future conditions. *Source: EXECUTIVE SUMMARY and chapter excerpts, SWITZERLAND Financial Sector Assessment, INTERNATIONAL MONETARY FUND.* ### EXECUTIVE SUMMARY __________________________________________________________________________ 7 ### EXECUTIVE SUMMARY ### Introduction and recent shocks - The Swiss financial system has navigated turbulent times since the 2019 FSAP, including the COVID 19 pandemic, geopolitical conflicts, and the collapse of Credit Suisse (CS) in 2023. - The government-assisted merger between UBS and CS entailed significant contingent fiscal liabilities, undermined the credibility of the Too-Big-To-Fail (TBTF) regime, and revealed gaps in supervision, resolution, and crisis management. ### Financial system environment and structural vulnerabilities - The financial sector operates in a generally favorable macro-financial environment but faces structural vulnerabilities: - Rising concentration in the financial system and associated risks of competitive distortion. - Periods of strong Swiss Franc appreciation that burden Swiss exporters and complicate monetary policy. - High household debt and large banks’ mortgage credit exposure. - High and rising house prices and housing affordability strains. - Weak banking system profitability driven by high operating costs and low interest rates. - Structural factors supporting high household mortgage debt include tax incentives, interaction with pension saving schemes, and lenient amortization requirements; policy choices should aim to counter further household debt creation and stretching house price valuations (see companion FSAP Technical Note on Macroprudential Policy and Real Estate Risks). - Real estate (RE) firms and associated commercial real estate activity should be monitored vigorously given market-based financing structures; more data collection is warranted. ### Bank solvency stress testing — findings and caveats - The bank solvency stress test indicates the Swiss banking system can withstand severe economic downturn scenarios: - Solvency ratios fall notably but remain above minimum requirements at the system level under severe demand and supply shock scenarios. - Capital depletion is more pronounced for asset and wealth management banks than for other banks. - Important caveats include high uncertainty regarding certain risk parameters for asset and wealth management banks. - Recommendation: Enhance data coverage for the asset and wealth management bank cluster and incorporate them into authorities’ stress tests and risk analyses. ### Bank liquidity risk — recent dynamics and stress test - Liquidity risk profiles changed materially since the COVID pandemic due to: - The “dash for cash” during the pandemic. - Directional turns in monetary policy since 2022. - The UBS-CS merger in 2023 causing notable deposit outflows and flows among banks. - Forward-looking liquidity stress tests identify a few weak banks, mostly from regional and asset and wealth management clusters, that would face liquidity strains under hypothetical severe liquidity withdrawal scenarios. - The liquidity shortfall is not negligible but is sufficiently contained from a system perspective. - Recommendation: Develop a bank liquidity stress test model to support analytical capability. ### Insurance sector — solvency and liquidity - The solvency stress test for six Swiss insurance groups suggests resilience: they can withstand severe shock scenarios, though solvency ratios fall significantly. - Key shock drivers: higher credit spreads, shocks on property holdings, and equity exposure. - Insurance transfer mechanisms limit the banking system impact from climate physical risks. - Insurers have safeguards against liquidity risk: - Access to the repo market (with SNB activity). - Contractual and legal restrictions to policy surrenders. - Large notional derivatives positions used to hedge CHF appreciation and, to a lesser degree, falls in interest rates. - The solvency stress scenario would have led to a net positive liquidity impact from derivatives holdings. - Recommendation: Supervisors should run regular liquidity stress tests for insurers. ### Pension funds - The pension fund sector appears robust despite market stress and increasing life expectancy. - Pension fund bankruptcies are rare and have mostly involved small institutions. - Gross insolvency payments by the guarantee fund have been low and stable. - Some pension funds use leeway in calculating funding positions (e.g., higher technical interest rates) which can overstate funding and keep fluctuation reserves lower than justified. - Recommendation: Close identified data gaps and develop regular data collection to enable market-wide horizontal and systemic risk analyses, particularly regarding asset exposures. ### Climate physical risk - Climate physical risks can affect the banking system, with cantonal and regional banks most exposed by various indicators. - An extreme flood scenario could impact property values and bank capital ratios, with a significantly higher effect than from increasing premiums due to higher average annual flood damages. - Insurance transfer mechanisms limit the effects on the banking system. - Recommendation: Collect more data from financial institutions to quantify climate-related risks and collaborate with the Federal Office of the Environment, building insurers, and other authorities. ### Cross-institutional monitoring and methodology enhancements - Data and monitoring gaps identified across institutions include mortgage details for non-SIB banks, banks’ trading and investment portfolios, banks’ foreign exposures, Lombard loan exposures, commercial real estate transaction data, and bilateral exposures across financial institution types. - Recommendations include developing micro data-based models for nonfinancial corporates and households, developing model structures for asset and wealth management banks, benchmarking provisioning practices, and ensuring comparability of pension funds’ funding position calculations. - Recommendation: FINMA and SNB should collaborate on reciprocal benchmarking of top-down and bottom-up stress test results and establish a common platform to assess market/sectoral developments and risks. ### Main recommendations (selected and verbatim where possible) - 1 Keep monitoring housing and mortgage market developments, including house price valuation, affordability strains and interest rate risk for households, and consider related macroprudential policies to contain systemic risk. - Timing: ST/C; Responsible: FINMA, SNB - 2 Monitor and collect more data pertaining to commercial real estate business, real estate (RE) firms’ borrowing from banks and capital markets, to assess their relevance and contribution to systemic risk and vulnerabilities vigorously and continuously. Also consider developing transaction-based price indices for commercial RE. - Timing: ST/C; Responsible: FINMA, SNB - 3 Compile bilateral exposure data involving all financial institution types, beyond banks, for authorities’ network and contagion analysis to achieve a fuller system coverage. - Timing: MT; Responsible: FINMA, SNB - 4 Collect more data on a regular basis to enhance the risk analysis and modeling, including related to mortgages for banks other than SIBs, banks’ trading and investment portfolios, banks’ foreign exposures in general, and Lombard loan exposures specifically. - Timing: MT; Responsible: FINMA, SNB - 5 For insurers, reduce dependence on ad-hoc data requests and establish more frequent regulatory reporting. - Timing: MT; Responsible: FINMA - 6 Develop a dataset and regularly collect data for pension funds to enable market-wide horizontal and systemic risk analyses, particularly regarding asset exposures. - Timing: MT; Responsible: OPSC - 7 Collect more data from financial institutions to quantify climate-related risks. The authorities should collaborate with the Federal Office of Environment, building insurers, other relevant authorities, and private sector entities to monitor such risks. - Timing: MT; Responsible: SNB, FINMA - 8 Develop a bank liquidity stress test model, to complement the monitoring based on data and the analysis of regulatory metrics as so far prioritized. - Timing: MT; Responsible: SNB, FINMA - 9 Keep developing micro data-based models for nonfinancial corporates and households and connect them to the bank solvency stress test model. - Timing: MT; Responsible: SNB - 10 Develop additional model structures pertaining to asset and wealth management banks, to include them in the stress test models and thereby support the analysis of their vulnerabilities and systemic relevance. - Timing: MT; Responsible: SNB, FINMA - 11 Benchmark and assess the banks’ provisioning practices for performing exposures. - Timing: MT; Responsible: FINMA, SNB - 12 Conduct liquidity-focused top-down stress tests and scenario analysis for insurers. - Timing: MT; Responsible: FINMA - 13 Ensure comparability of the calculation of pension funds’ funding position. Provide cantonal supervisors with the necessary authority and instruments to intervene. - Timing: MT; Responsible: OPSC, cantonal supervisors, Federal Council - 14 Consider providing more information about the macro-financial scenarios and bank solvency stress test results in SNB’s financial stability report. - Timing: MT; Responsible: SNB - 15 FINMA and the SNB should collaborate regarding the reciprocal benchmarking of top-down and bottom-up stress test results for banks. - Timing: MT; Responsible: FINMA, SNB - 16 Establish a common platform to assess and discuss market/sectoral developments and potential risks to financial stability, the findings of which should feed into the Steering Committee/Standing Committee for financial stability discussions. - Timing: MT; Responsible: OPSC, Federal Council, cantonal supervisors, FINMA, SNB *Source: EXECUTIVE SUMMARY, SWITZERLAND, INTERNATIONAL MONETARY FUND.* ### INTRODUCTION ### INTRODUCTION ### Purpose and scope of the SRA - The note presents the findings and recommendations of the systemic risk analysis (SRA) conducted for the Swiss financial system as part of the Switzerland 2025 FSAP. - Objective: identify sources of vulnerabilities and systemic risk facing the Swiss financial sector, inform policy advice, strengthen system resilience to external adverse shocks, and detect and counter endogenously building vulnerabilities. ### Methodology and exercises - Stress test exercises covered banks and insurance firms. - Risks evaluated at individual firm level and system perspective using: - Models for solvency and liquidity for banks and insurers. - Solvency-liquidity feedback for banks. - Inputs from nonfinancial corporates and households risk analysis into banks’ solvency analysis. - Real estate sector vulnerability analysis to inform stress test scenarios. - Climate risk analysis assessing impacts on insurers and banking system solvency under adverse physical risk materialization scenarios. - Emphasis on structural models over econometric models due to Switzerland’s long period of low and negative policy interest rates and very low default rates for mortgages. - Bank-level quantitative outputs combined to compute vulnerability rankings to identify entities that are both vulnerable and impactful. ### High-level findings on financial system structure - Size of financial system at end-2023: about 1,000 percent relative to Swiss nominal GDP. - Split: about 600 percent for NBFIs and 400 percent for the banking system. - Four largest banks (designated SIBs: one GSIB and three DSIBs) represent 264 percent of GDP and 65 percent of banking system assets. - Consolidation trends: - Number of banks fell from 275 in 2014 to 213 in 2024 (22 percent decline). - Banking system assets shrank from 500 percent relative to GDP in 2021 to 400 percent in 2024. - UBS–CS merger in March 2023 materially increased concentration. - Banking system clusters (end-2024): large banks (SIBs), cantonal banks, asset and wealth management banks (stock exchange banks), regional banks, and a residual "other" cluster. - Cantonal banks: cantons hold more than one third of shareholder voting rights for cantonal banks; various forms of cantonal guarantees exist for all cantonal banks except one. - Insurance sector: - Annual insurance premium flows ≈ 60 percent of Swiss GDP. - Market concentration: five largest life insurers account for 80 percent of the life market; five largest non-life insurers account for 67 percent of the non-life market. - Swiss Re: second largest reinsurer in the world, earning nearly all premium income outside Switzerland. - Pension funds: - Assets around 140 percent relative to GDP. - More than 1,300 occupational pension institutions; top 14 percent of largest entities represent 82 percent of active members and total pension assets. - Defined-contribution regimes represent almost 90 percent of pension liabilities. - NBFI remainder: - Investment funds and OFIs each represent about 160 percent in total assets relative to GDP. - Data coverage limitation: financial accounts statistics only account for Swiss portion of sector, making total balance sheet size of Swiss NBFIs difficult to estimate. - Since 2022, authorities collected data on exposures, leverage, liquidity, and counterparty risks of investment funds. - Some NBFI categories lack oversight or reporting, complicating financial stability assessment. - Financial market infrastructures: - SIX Group AG provides trade and post-trade services, operating SIX Stock Exchange, RTGS, CCP, CSD, SSS, SDX, and SIX Structured Products Exchange. - SIX Stock Exchange free float market capitalization ≈ CHF 1.8 trillion (225 percent of GDP, at end-2024). - Digital assets and fintech: - SDX went live in 2021 following DLT legal framework introduction. - Project Helvetia advanced to a pilot phase exploring a wholesale CBDC for settling tokenized financial assets. - Fintech license allows acceptance of public deposits up to CHF 100 million in crypto assets under specified conditions. ### Macro-financial environment - Economic backdrop: - GDP growth steady and supported by exports. - Inflation stable and low, with occasional deflation over past two decades. - Switzerland experienced only modest inflation pressures following the pandemic and the Russia-Ukraine war. - Foreign/imported component of inflation ≈ 24 percent according to SNB estimates. - Interest rates and exchange rate dynamics: - Policy and market rates were low and negative for long, then began to rise since Jan-2022. - Rate hike cycle began in June 2022; rate increases compressed and inverted the yield curve during end-2022 to end-2024. - CHF exhibits a structural appreciation trend, weighing on export competitiveness. - Fiscal and external strengths and vulnerabilities: - Strong fiscal fundamentals: general government gross debt to GDP at 25 percent in 2024; total public debt at 32 percent of GDP. - Openness to financial innovation and safe haven perception. - Macro-financial vulnerabilities: anticipated financing gap of the pension system, sensitivity to foreign demand compounded by CHF appreciation, and elevated nonfinancial private sector debt including household debt. - Nonfinancial private sector debt: - Nonfinancial private sector debt high at 267 percent relative to nominal GDP in 2024. - SNB balance sheet sensitivities: - SNB FX holdings are sizeable and unhedged; FX moves influence SNB profits and equity. - Higher bank reserve requirements enacted from 2.5 percent to 4 percent since July 2024. - In 2024, SNB profits were supported by FX and equity gains, a stronger USD, and valuation gains from gold. - Regulatory cooperation: - Mutual recognition agreement between the UK and Switzerland concluded in December 2023 to deepen financial services provision and jointly recognize regulatory regimes. ### Banking system profitability, solvency, and liquidity — key metrics and trends - Profitability: - Banking system ROA was 0.4 percent in 2024 (just at the average of the past ten years). - One-off ROA in 2023 was 0.9 percent driven by UBS’s takeover of CS and consequential valuation gain. - Profitability is heterogeneous: SIBs and regional banks less profitable than asset/wealth management banks. - Large share of income for big banks derives from fees and commissions. - Solvency and asset quality: - Capital ratios at system level rest at moderate levels in cross-country comparison. - Gap between actual and required capital ratios is smaller for SIBs and cantonal banks; asset and wealth management banks have higher buffers. - NPL ratios are structurally low, reflecting historically low default rates. - Risk weight densities range between 40-50 percent. - Liquidity: - Liquidity risk profiles have been in flux since the pandemic due to the “dash for cash,” SNB policy shift since 2022, and the UBS–CS merger in 2023. - Share of banks’ reserves in total assets fell from 30 percent to 25 percent around September 2022 and to 20 percent until end-2024. - Interest rate increases induced reallocation from sight deposits to term deposits, which reduces liquidity risk from that perspective. - LCRs at system level rank well in cross-country comparison but are heterogeneous across bank clusters and volatile. - NSFRs show asset and wealth management banks have stronger liquidity profiles than other banks. - Liquidity ordinance amended in July 2022; higher liquidity requirements foreseen for SIBs. - A review of stricter liquidity requirements expected by end-2026; discussion under way on allowing temporary LCR below 100 percent in stress if justified by contingency plans. - Banking system balance sheet composition (end-2019 to end-2024 highlights): - Mortgage lending grew to represent 37 percent of total assets at end-2024 (30 percent at end-2019; mortgages constitute 66 percent of the loan book). - Drop in total assets late-2022 to 2024 driven primarily by falling cash/reserves and interbank lending, alongside contracting non-mortgage credit. - Customer deposits, interbank borrowing, own bond issuances, and mortgage bonds are main liabilities; drop in liabilities late-2022 to 2024 largely due to negative deposit growth. - Monetary developments: - Currency in circulation ≈ 6 percent and all forms of deposits ≈ 94 percent of total money. - Negative money growth during 2022-24 accompanied by shift from sight to term deposits. - Term deposit rate shown pertains to 12-month term deposit contracts and is a mean across Swiss banks. *Source: IMF staff.* ### 22. House price dynamics and valuation analyses suggest clear signs of overvaluation. ### 22. House price dynamics and valuation analyses suggest clear signs of overvaluation ### House price valuations and model estimates - SNB estimates that house price overvaluation ranges between 15–40 percent. - Price-to-income and price-to-rent ratios indicate house prices rest by about 30 percent above historical averages. - Model-based analysis suggests point estimates between 15-25 percent, while upper confidence bands reach up to 35 percent. - A conservative estimate between the point and upper bound estimates was used to inform house price shocks in the adverse macro-financial scenarios. ### Structural drivers of house prices and mortgage debt - Structural factors contributing to house price growth include immigration and scarcity of suitable land for construction. - Factors that support high household mortgage debt: - Tax incentives, interplay with pension saving schemes, and lenient amortization requirements create an environment where maintaining high levels of mortgage debt is financially rational and institutionally supported. - Five specific supporting factors: 1. At origination, mandatory pension savings (Pillar 2) and voluntary pension savings (Pillar 3) can be pledged for mortgage borrowing, allowing higher leverage. 2. At origination, 10 percent of the lending value of the property must be covered by the borrower‘s own funds; the remaining part of the down payment can be covered by depleting Pillar 2 or 3 savings (for Pillar 2, up to 100 percent of the savings until age 50; the higher of up to half of it or the amount at 50 after age 50). 3. The option to have up to two thirds of total lending value not amortizing. 4. Pillar 3 savings can be used for “indirect amortization,” i.e., borrowers save in a Pillar 3 scheme to repay the mortgage at maturity while retaining tax benefits from interest deductibility and generating interest income and capital gains from Pillar 3 investments. 5. Interest on mortgage debt is tax deductible against income. - Non-amortizing mortgages are particularly prevalent among lower-income borrowers: the share of such non-amortizing debt for borrowers in the lowest income buckets reaches up to 70 percent. - The first mortgage, up to two thirds of the lending value, does not need to be amortized; the second mortgage (exceeding two thirds) is to be repaid within 15 years or by retirement (whichever comes first). - Indirect amortization via Pillar 3 resembles a “bullet loan” scheme though Pillar 3 is more institutionalized and savings are pledged explicitly for the second mortgage. ### Housing market metrics and trends - Homeownership rates currently stand at about 40-45 percent. - House price growth in Switzerland over the past five years amounted to about 11 percent in real terms. - Vacancy rates are low in cross-country comparison, pointing to supply side strains. - Construction costs are stabilizing, while actual construction activity keeps contracting. - Mortgage credit constitutes a large share in total credit; real estate lending represents a sizeable share of bank lending (>70 percent of the loan book). - Variable rate mortgage shares have risen to 20 percent. - Housing costs for owners are comparably low due to the elevated share of non-amortizing mortgages. - Loan-to-value (LTV) at origination is highest for owner-occupied housing; LTVs for CRE are most volatile. - Debt service-to-income (DSTI) ratios are higher for buy-to-let business and follow interest rate developments. - Loan-to-income ratios are higher for private households’ buy-to-let activity. ### Risks, vulnerabilities, and market structure concerns - Banking system exposures and concentrations: - For some bank clusters (e.g., various cantonal and regional banks), CRE lending reaches 25-30 percent of their loan books (including credit to firms active in the RRE market). - Various sizeable RE firms in Switzerland borrow largely unsecured from banks, resulting in a mostly unencumbered asset base for them and more limited and uncertain recourse to RE firms’ assets in case of default. - Financial sector exposure composition: - Real estate-related investments account for 31 percent of life insurers’ investment portfolios. - Pension funds hold 23 percent of their assets in real estate-related investments. - Systemic concentration risk: - UBS, after takeover of CS, is the world’s largest G-SIB relative to domestic GDP and provides critical functions for the Swiss financial system; it outranks peer G-SIBs in metrics of complexity, cross-jurisdictional activities, and interconnectedness. - Macroeconomic vulnerabilities: - The Swiss economy is vulnerable to external demand shocks and global supply shocks that could raise interest rates, burden variable rate borrowers, and depress security valuations. - A housing market downturn could spike loss given default (LGD) parameters, increase provisioning needs, and pressure bank solvency; default probabilities may rise particularly in the CRE segment. ### Macro-financial scenarios and stress-test framing (2025–27) - Scenario horizon: three-year period from 2025–27. - Baseline scenario is aligned with the IMF WEO (an intermittent IMF-internal one as of February 2025). - Two adverse scenarios were developed: - Adverse (D): demand shock-dominated disinflationary scenario (involves a reduction in policy rates). - Adverse (S): supply shock-dominated inflationary scenario (entails an increase in monetary policy rates to counter inflationary pressures). - Both adverse scenarios include a global component assuming deepening geoeconomic fragmentation and materialization of global risks. - Housing markets are assumed to experience a downward correction under both adverse scenarios, with severity informed by the disequilibrium analysis. - Scenario severity metrics (business cycle-state dependent): - Standard deviation multiples based on real GDP growth relative to historical mean growth for year 1 and cumulative 2 years: - Demand shock scenario: 2.4x and 2.3x. - Supply shock scenario: 1.7x and 1.9x. - The supply shock scenario is deemed adequately severe in a multivariate sense because inflation and interest rates experience upward pressure. ### Policy implications and recommended monitoring - The non-amortization and indirect amortization schemes imply risks including: - Structurally high household indebtedness and continuous structural upward pressure on housing demand and valuations. - Credit risk when principal comes due, especially if Pillar 3 investments include higher shares of risky securities whose value may drop. - Combined high household leverage and concentrated mortgage lending raise the risk of a housing-led financial crisis and procyclicality. - Authorities should carefully monitor associated vulnerabilities and consider preventive policies that contain household indebtedness and debt service burden well in advance. - Further analysis recommended: - Analyze RE firms’ use of secured vs. unsecured bank borrowing by RE firm size, using RE firms’ financial reports and supervisory reporting from FINMA. *Source: IMF staff analysis, “House price dynamics and valuation analyses suggest clear signs of overvaluation.”* ### 32. The bank stress test sample comprised SIBs, cantonal banks, private banks engaged in ### 32. The bank stress test sample comprised SIBs, cantonal banks, private banks engaged in ### Sample composition and coverage - The solvency and liquidity stress testing and interconnectedness analysis used a common sample of 92 (112) banks at consolidated (solo) level, covering 93 percent of Swiss banking system assets at end-2024. - Sample breakdown: - Four SIBs (one of which is the aggregate Raiffeisen bank holding) - 23 cantonal banks (a 24th being part of the SIBs cluster) - 12 asset and wealth management banks - 42 regional banks including cooperatives - 11 other banks ### Solvency stress test model architecture - The solvency stress test model combines structural and econometric elements developed for Swiss banks. - Credit risk setup: - Seven portfolios defined with PD and LGD models (Table 2). - Econometric PD components for NFCs, CRE firms, banks, and NBFIs (Appendix E). - PD estimation uses data for 432 Swiss firms from Moody’s Credit Edge for estimated PDs, leverage, and other metrics. - A Bayesian Model Averaging (BMA) methodology was used to relate PDs to macro-financial variables for scenario-conditional forecasting. - Household mortgage portfolios: - A micro-macro simulation model (IDHBS+) based on confidential micro data from the Swiss Federal Statistics Office (FSO)’s SILC survey and supplementary data covering 8,900 households and 19,229 household members. - Structural micro-data model employed because no historical data for household default rates or LGDs are available to the Swiss authorities; historical default rates and NPL formation were very low. - Uses of the household simulation model: - (1) Compute scenario conditional PD paths for retail mortgage portfolios and consumer credit, and LGDs for mortgage segment, as input to the bank solvency stress test; - (2) Conduct counterfactual simulations in the macroprudential policy context; - (3) Build in a climate model overlay to assess how household risk parameters change under physical climate risk materialization scenarios. ### Key model components (as summarized in Table 2) - Interest expense and income: panel econometric bridge equations linking banks’ cost of funding and interest income to market/policy rates, banks’ own capital ratios, and for cross-border banks, rates from other jurisdictions. - Credit risk PD/LGD modeling: - NFC excl. CRE: BMA for PDs from Moody’s Credit Edge; involves 432 firms. - CRE: link to commercial property prices under scenario at bank-portfolio level. - Household mortgages: Structural micro-macro simulation model (IDHBS+), link to residential property prices at micro level. - Consumer credit, banks, NBFIs, foreign RE-backed & other: Frye-Jacobs and scaling approaches as specified. - Fee & commission income: panel BMA relating income/expense ratios to macro-financial conditions; FCI for asset & wealth management taken as ratio to off-balance sheet assets under management. - Market risk: Bonds in HFT and AFS revalued using modified duration, conditional on scenario interest rate trajectories; participations and equity revalued per equity price trajectories; SNB models used for SIBs; only HFT and AFS portfolios revalued (HTM not marked-to-market in main run). - RWA and credit growth: IRB RWs modeled with IRB formulas; STA portfolios use migration effects from performing to nonperforming to drive RW changes. - Non-interest expenses: residual P&L item up to net income before tax; ratio to total assets gross of provisions held constant. ### Econometric models for income and expenses - Structurally informed econometric models set up for: - Interest income: pass-through from policy and market rates, banks’ cost of funding, and macro-financial variables driving credit spread component. - Interest expense: pass-through from policy and market rates and feedback from bank solvency to cost of funding. - Fee and commission income: panel BMA methodology per bank cluster. - Data underpinning these models cover 92 Swiss banks (consolidated group level) representing 93 percent of banking system assets at end-2024; flow measures are 4-quarter trailing sums and denominators are 4-quarter trailing averages. Last observations are for 2024Q4. ### Market risk module and portfolio data - Bank-level exposure data processed into 13 segments (participations, money market investments, bonds, shares; domestic vs foreign; sovereign/financial/NFC bonds). - Modified duration formula used to revalue bond portfolios conditional on scenario interest rate shifts; SIBs use SNB’s more detailed market-risk models. - Only HFT and AFS portfolios revalued in the baseline revaluation; a sensitivity analysis revalued HTM portfolios. ### Corporate and household sector model inputs and scenario responses - Scenario horizon for PDs: 2025(Q1)-2027(Q4). Adverse (D) and supply (S) denote demand and supply shock scenarios. - PD trajectory findings: - Portfolios other than household mortgages and consumer credit: demand shock scenario causes more pronounced PD hikes because GDP growth, unemployment, and house prices are more adverse in the demand than in the supply shock. - Supply shock scenario is more detrimental for household mortgages and consumer credit due to rising interest rates and households’ elevated interest rate sensitivity. - Household mortgage PDs are not directly reactive to residential house price drops in the structural IDHBS model because full recourse reduces incentives for strategic default in Switzerland. - LGD trajectory findings: - Household mortgage LGDs rise from about 10 percent to 30 percent under the demand shock scenario. - CRE portfolio LGDs rise from 15 percent to more than 50 percent at the peak. - LGDs for NFCs, banks, NBFIs, and consumer credit modeled using Frye-Jacobs; LGDs for CRE and household mortgages linked to commercial and residential property prices in the scenario. ### Solvency stress test results — aggregate and cluster outcomes - Aggregate system resilience: - Aggregate CET1 capital ratio of 92 banks moves from 17 percent at end-2024 to 10.8 percent at the trough. - The capital ratio of the aggregate sample of 92 banks drops by 6.2 p.p. to the low point in Year 2. - Cluster-specific impacts (end-2024 starting points and drops): - Cantonal banks’ solvency least affected: -4.3 pp. - Asset managers: starting point CET1 20 percent and drop -13.6 pp. - SIBs aggregate CET1 ratio drops -6.3 pp; all SIBs remain above hurdle rates. - Drivers of capital decline (supply shock, up to third year end-2027, bank-system contributions in 3-year cumulative terms): - Loan losses: -4 pp difference between adverse and baseline. - Fee and commission income: -2.2 pp. - Trading income: -1.7 pp. - Net interest income: -1.4 pp. - Heterogeneity of drivers across clusters: - Loan losses mainly driven by deteriorating retail mortgage portfolios with strong impact on regional banks, cantonal banks, and other smaller banks. - Rising interest rates increase funding costs, higher loan losses from variable-rate borrowers, and bond valuation losses (smaller effect where hedging is notable). - Lombard loan losses relevant for asset and wealth managers and SIBs. - Fee and commission income and trading impacts pronounced for asset and wealth managers, but results are subject to sizeable model and parameter uncertainties for that cluster. ### Shortfalls and macroeconomic significance - Hurdle rates in adverse scenarios: regulatory minimum capital requirements (Pillar 1) plus Pillar 2 requirements and systemic surcharges for the four SIBs; CCoB and CCyB allowed to be consumed. - Outcomes: - Five of 92 banks fall below the CET1 hurdle rate by the third year of the supply shock scenario. - Six banks fall short of hurdle rates on total capital ratios. - These banks’ combined assets represent 4.6 percent of banking system assets at end-2024. - Their combined CET1 capital shortfall amounts to 0.4 percent of GDP under the supply shock scenario’s third year (and 0.04 percent of GDP under the demand shock scenario). - Model uncertainty caveats (noted especially for asset and wealth management banks): - Uncertainty regarding starting-point risk parameters for sizeable Lombard loan exposures (e.g., LGDs). - Extent of hedging of trading and investment portfolios not fully observed. - Specific dynamics of fee and commission income for asset/wealth management banks. ### Sensitivity analyses - HTM revaluation sensitivity: - Optionally revaluing banks’ HTM bond holdings yields economically relevant capital impacts, most visible under the supply shock scenario due to rising short- and long-term interest rates. - Capital impacts for cantonal and regional banks from HTM revaluation amount to -1.5 and -0.9 p.p., respectively. - Impacts are visible even for clusters with smaller investment portfolios because much is held under HTM and with longer durations. - Solvency-to-cost-of-funding feedback sensitivity: - Switching off the solvency-to-cost-of-funding feedback reduces adverse capital impacts for some clusters. - The feedback quantitatively explains between 40-50 bps of the adverse-baseline gap of capital ratios at the low point for asset and wealth management banks and regional banks. - Other sensitivity notes: - Hedging and bottom-up SNB/UBS model features (e.g., delta/gamma/correlation risks) influence market risk assessments and were captured for SIBs via SNB models and UBS bottom-up calculations where applicable. _Italic: Source: IMF staff analysis and models as presented in the chapter excerpt._ ### 46. A cash flow-based liquidity stress-test model was employed. It operates on banks’ ### 1cheea2025011-source-pdf - 46. A cash flow-based liquidity stress-test model was employed. It operates on banks’ ### Liquidity stress-test model design - The model operates on banks’ individual, detailed maturity profiles for different types of liabilities, coupled with assumptions for runoff rates for the maturing portion of these liabilities, and their assumed contractual term structure for the portion that rolls over. - Counterbalancing capacity includes all financial assets that are not loans and unencumbered, including: - cash and reserves at banks’ central bank account, - deposits in other banks (i.e., interbank claims), - reverse repo lending, - bond and equity holdings. - Bond and equity holdings are optionally revalued in line with the market risk shocks as used in the solvency stress test. - The model accounts for a reverse link back to solvency: HTM bond sales (sorted last in a sales hierarchy) can cause losses when they need to be sold to honor liquidity outflows. - The model captured the collateral add-back mechanism related to secured funding, while accounting for the drop in value of that collateral in line with the market risk model. ### Run-off, drawdown, and valuation assumptions - Run-off rates for the portion of liabilities that mature within 30 days (cumulative runoff rates) were assumed to be: - 10 percent for household sight deposits, - 30 percent for household term deposits, - between 30 and 40 percent for deposits from NFCs, banks, NBFIs, and bonds, - 15-25 percent for secured funding. - Banks’ open committed credit lines were assumed to be drawn by 20 percent within the first four weeks. - The run-off rates were informed by regulatory LCR stress assumptions as well as the recent run-off dynamics of CS. - An additional simulation input included valuation changes for bonds and equity exposures from the market risk module of the solvency stress test, applying the demand and supply shock scenarios with their different market stresses. - Haircuts and valuation drops for collateral under secured funding were applied in line with the market risk model. ### Banking system stress-test results - Sample: 92 banks. - Under the baseline liquidity simulation: - 10 banks were estimated to experience a liquidity shortfall before 30 days; all 10 are regional and other banks. - The liquidity shortfall of the failing banks accumulates to 0.15 percent of the initial liquid assets of all 92 banks. - Under the most severe scenarios with maximal stress on valuations of bonds and equity holdings (up to a six-month horizon): - Up to 19 banks were identified to face a liquidity shortfall, most of which are small regional banks. - This more severe stress was primarily used to rank banks in relative terms regarding their risk profiles rather than to derive “absolute” stress resilience. - System-level liquidity metrics (end-2024 observations, by bank cluster): - All banks satisfy LCR requirements, but Cantonal, regional, and other banks have lower LCRs than other bank clusters. - NSFRs rest at comfortable levels for the median of the banks in all clusters; they are higher for asset/wealth managers. - The cumulative funding gaps relative to banks’ initial liquid assets are most negative for regional banks. - Unencumbered liquid assets (reserves, bonds) as a percent of total liabilities are lowest for regional and other banks. - Cash and reserves relative to total liabilities are lower (at the median) for asset managers, regional, and other banks. - Off-balance sheet commitments (e.g., credit lines and guarantees) are highest for SIBs. - Secured funding (esp. via covered bonds) is sizeable across clusters; household term deposit liabilities are comparably most sizeable for regional banks. ### HTM bond sales and feedback to solvency - Feedback from HTM bond sales to bank solvency was found to be limited in general. - Under the supply shock extension where HTM bonds lose value at the outset amid rising interest rates: - At the 6-month horizon, 10 banks would face feedback of less negative than -0.5 p.p. shift for their solvency ratios. - Five banks would face feedback more negative than -1.5 p.p. - These affected banks include banks from the regional bank cluster and the asset/wealth managers. - Numerous banks avert an outright liquidity shortfall by selling the bonds in the simulations. - An alternative interpretation: rather than selling HTM bonds, banks could use them for repo borrowing from the central bank. ### Insurance sector—scope and characteristics - The insurance sector stress test: - Run on a consolidated group basis using end-2024 balance sheets as a baseline. - Included six Swiss insurance groups, accounting for about two thirds of the Swiss life and non-life insurance market and 70 percent of the reinsurance sector in Switzerland. - One group is primarily domestic; five are designated as IAIGs with between around 40 and 90 percent of premia written outside their home market. - For two groups, the reporting currency is USD. - Investment composition and quality: - Insurers’ investments account for 60 percent of their balance sheet; unit linked business-related assets add another 25 percent. - More than half of investments relate to government and corporate bonds. - Direct property investments plus mortgages and real estate funds represent more than a fifth of insurers’ own investments. - More than 90 percent of bonds are of investment grade; around 56 percent are of AA rating or higher (credit quality 1 or better). - Cash levels are above two percent of total assets. - Mortgage exposure is high in a European context but with low LTV (around 60 percent LTV at origination on average) and “DSTI-type” metrics below 40 percent where data is available. - Geographic and currency exposures: - Most financial investments are outside the home country; only 19 percent of equities and 15 percent of bonds are held in local currency. - Currency exposures on the asset side are largely aligned with those on the liability side; remaining FX risk is partially hedged via currency derivatives. ### Insurance stress-test methodology and scenarios - Macro-financial scenarios used for the banking sector tests also served as the basis for the insurance stress test with adjustments to be insurer-applicable. - For the insurance tests: - All shocks were assumed to occur at the beginning of the first year (instantaneous shock approach). - Market shocks were front-loaded so maximum drawdown during the project horizon is realized immediately after the reference date (31 December 2024). - Reactive management actions were requested and assessed where they relate to non-discretionary arrangements in place at end-December 2024; reactive management actions were modelled in the BU exercise but not in the TD exercise. - Market risk stresses were defined granularly and include shocks to: - the risk-free interest rate, - equity prices, - real estate prices, - default rates of mortgage loans, - credit spreads of corporate and sovereign bonds, - exchange rates. - The Swiss Solvency Test (SST) is the basis for the insurance stress test; main output is effect on risk-bearing capital and coverage of the target capital. - BU approach: insurers could use FINMA-approved full internal models where applicable. - TD approach: applied shocks to investment assets and insurance liabilities, applied haircuts to market values, revalued fixed income assets with stressed term structures by currency; TD did not include separate SST coverage ratio calculations nor look-through to unit-linked business. - Three-year projections (baseline and adverse) were requested for premia, claims, lapse rates, investment returns, net earnings, insurance liabilities, risk-bearing capital, and target capital; instantaneous shock assumption implies market value of investments stayed constant after shock and recovery in solvency would be driven by underwriting and recurring investment income. ### Insurance stress-test results - BU results: - Median SST ratio drops from 255 percent before stress to 219 percent after stress. - No insurer fell below the regulatory minimum in the BU calculations. - Only two insurers applied reactive management actions; these actions increased SST ratios by 15 and 20 percentage points for those companies. - TD / whole-sample aggregation: - For the whole sample, the SST ratio fell by more than 40 percentage points (size-weighted aggregate), from a baseline of 236 percent. - Main driver of the fall was reduction in risk-bearing capital after stress. - Balance sheet shrank by CHF 180 billion post-stress. - Risk-bearing capital was almost halved. - New target capital falls by 32 percent due to lower asset and liability exposures post-stress. - Average assets over liabilities ratio fell from 117 percent to 110 percent. - Corresponding loss in equity (excess of assets over liabilities) was nearly CHF 60 billion. - Overall conclusion: - The industry can withstand severe price shocks, despite vulnerabilities on the asset side; market shocks lead to significant valuation losses and materially reduce risk-bearing capital and SST ratios, though regulatory minima were not breached in the BU sample. *Italic: IMF staff calculations and analysis based on FINMA, SNB, and company submissions as presented in the source content.* ### 65. Asset valuation losses due to significant interest rate hikes are compensated for by ### 65. Asset valuation losses due to significant interest rate hikes are compensated for by ### Interest rate shock, valuation, and solvency - Swiss insurers have an average negative duration gap of around two years. - Net valuation effect of the rise in interest rates is positive and: - bolsters risk bearing capital; and - tends to lower target capital because capital requirements are applied to a lower base. - Higher interest rates may negatively impact insurers’ income statement and liquidity position. - Insurers’ projections are heterogeneous; higher lapses in the short term are conceivable (similar to the pandemic period). - Some insurers may face liquidity losses from margin calls due to higher short positions in interest rate derivatives. - The largest net impact on excess of assets over liabilities in the adverse scenario stems from fixed income related spread risk, followed by property and equity risk. ### Scenario: fall in Swiss interest rates (other shocks unchanged) - Impact is moderate due to short asset and liability durations and strong ALM positions. - Liabilities would increase slightly more than assets, leading to a net negative valuation impact. - Close asset–liability matching and hedging mitigate the impact. - Cash flow analysis: after five years the difference between net liabilities and net assets converges quickly to zero, highlighting good matching. - Lower guaranteed interest rates compared with the start of the prolonged low-interest rate environment reduce the potential impact of another low-rate period. ### Currency risk - Strengthening of the CHF against a basket of currencies has a small net negative impact, particularly for institutions reporting in CHF. - Foreign currency asset exposures are in most cases largely aligned with insurance business in respective countries; currency risk arises from residual differences and is mitigated by hedging against a stronger home currency. ### Spread, property, and equity risk - Spread risk and property risk, followed by equity risk, have the largest net impact on excess of assets over liabilities. - Real estate exposures (and propagation across asset types) amplify sector exposure: - Direct real estate represents around 15 percent of total investments. - Mortgages represent another six percent. - Real estate funds around one percent (three percent of their exposure to collective investment schemes). - Insurers may also hold equities and bonds of real estate corporations. ### Solvency projections and recovery - Insurers’ own three-year projections after the event indicate an improvement of the solvency position but not a full recovery. - Scenario led to an average fall in the SST ratio of around 60 percent and is expected to recover around 30 percentage points by end 2027. - Solid underwriting performance and investment returns improve profitability and technical results but do not fully restore solvency within three years. ### Liquidity risks and mitigants - Insurers are exposed to liquidity risk; prominent drivers are margin calls on derivatives and policyholder surrenders. - Swiss insurers use FX derivatives for currency hedging and interest rate derivatives to address duration mismatches. - Swiss insurers have a significant share of lapsable policies, though surrender values may be at a substantial discount. - Access to liquid repo markets and SNB standing facilities (intra-day liquidity and liquidity-shortage financing facility) are important mitigants for larger institutions. - Larger insurers participate in a bilateral repo market operated by SIX Swiss exchange; the SNB is active on this platform. - Contractual and legal restrictions on surrender values: - Surrender value must be paid within three months unless FINMA extends in limited circumstances. - Insurers may deduct for interest rate risk and partially claw back sales costs (up to five percent of the premium sum amortized over contract duration). - Combined deductions cannot exceed one third of the surrender value. - Severe restrictions apply when premia are tax-deductible; tax disincentives also apply to lump-sum payouts instead of annuitization. - Liquidity analysis using the solvency stress test scenario (instantaneous shock) focused on mass lapses and variation margin calls; no separate liquidity stress test was conducted. - Scenario is net positive for insurers’ derivative books: higher margin calls from increasing interest rates are offset by additional liquidity from currency derivatives hedging an increase in CHF. - Further mitigants: - Low asset encumbrance levels. - Unpledged cash to total assets ratios of more than two percent on average. - Access to the repo market. - High shares of high-quality liquid assets buffer mass lapse events: - Best estimate value in aggregate exceeds total surrender values by a considerable margin. - At the group level, insurers have sufficient liquid funds to withstand significant redemptions, particularly given deferred payouts. - Non-immediate payout of surrender values allows servicing via investment income and premia. - On average, around nine percent of fixed income assets mature every year. - Assuming a quarter of these assets mature within the first trimester and an average annual coupon of around two percent, quarterly cash inflows would represent around ten percent of total surrender values. ### Cyber risk (sensitivity analysis) - Cyber sensitivity analysis assessed two underwriting loss scenarios based on end-2023 data; operational cyber losses (direct losses incurred by insurers) were not covered. - Scenario 1: One-week outage of the largest cloud service provider (CSP) after a cyber-attack. - Scenario 2: State-based actor exploits a software update mechanism vulnerability to deliver malicious payload (Petya/WannaCry-type). - Cyber insurance has grown substantially since 2020 but remains a small share of total business: - Total premium volume in 2020 was around CHF 450 million, growing by more than 130 percent over the following three years. - Cyber-related premia remain less than one percent of total business. - Gross cyber claims can significantly exceed premium income but direct impact on SST ratio remains small: - Ransomware attack scenario: average fall in SST ratio of five percentage points. - CSP outage scenario: average fall in SST ratio of three percentage points. - Insurers assume no silent cyber losses; explicit exclusion of silent cyber exposure is an important mitigant. - Significant cyber-specific loss ratios indicate potential for larger losses if cyber underwriting continues to grow. ### Pension fund sector: systemic relevance and structure - Systemic risk of pension funds is generally low, but sector size and capital market footprint warrant scrutiny. - Factors relevant to systemic footprint include concentration and size, fund size relative to sponsor, plan type, home bias, herding, derivatives and leverage, access to funds by individuals, and flexibility of restoration measures. - Current structure dampens systemic footprint: - Sector is large but fragmented into many, often very small, institutions. - Asset allocation is geographically diverse except for real estate and mortgages (mostly domestic). - Restricted early access to funds limits liquidity concerns. - Restoration measures are broad and flexible and can be stretched over several years. - Rules for derivatives are comparatively strict and forbid their use for buildup of leverage. - Market consolidation and concentration risks: - Consolidation predominantly affects small single-employer funds; number of pension funds fell by 101 to 1,355 (91 relate to public-sector employers). - Ten years earlier there were 1,845 pension funds. - Since the FSAP in 2019, total asset size has grown by 12 percent, mostly due to a 10 percent higher number of insurants. - Number of active members has grown by three quarters of a million to 4.8 million. - The largest 14 percent of the sector (institutions with total asset size of more than CHF one billion) represent 82 percent of active members and total pension assets. - Pension funds with balance sheet size of less than CHF 100 million still cover more than 44 percent of all institutions but less than two percent of all active members and total pension assets. - Collective and joint foundations: - Represent 17 percent of all institutions. - Cover more than half of all assets and three quarters of active members. - Three fifths of retirees receive their pension from a collective or joint foundation. - Size metrics: - Second Pillar pension sector size: CHF 1.2 trillion or 152 percent of GDP by end-2023. - For comparison: Swiss banking and insurance sectors are about 400 and 60 percent of GDP, respectively. - Pension funds have sizable domestic real estate holdings: CHF 217bn compared to CHF 67bn for the insurance sector. *Source: IMF staff calculations and company submissions as presented in the original IMF chapter.* ### 89. Semi-autonomous solutions are gaining in importance, as smaller pension funds ### 89. Semi-autonomous solutions are gaining in importance, as smaller pension funds increasingly opt for partial insurance covering the investment risk ### Market structure and scheme types - Semi-autonomous solutions are increasingly used by smaller pension funds for partial insurance of investment risk. - The dominant solution remains the autonomous pension fund, which covers both investment and biometric risks: - Autonomous funds represent 35 percent of all pension funds and around 70 percent of respective liabilities. - Defined contribution (DC) regimes dominate the Swiss market: - DC schemes represent almost 90 percent of all pension liabilities. - Defined benefits (DB) schemes are concentrated in pension funds with a state guarantee, where DBs represent more than 40 percent of pension liabilities but only seven institutions. - Swiss DC schemes are hybrid cash balance plans: - Insured members bear market risk, but exposure is floored through a minimum guaranteed interest rate. - The conversion rate guarantees a fixed percentage of annual pension payout that cannot be reduced after retirement. ### Asset allocation, exposures, and market risk - Asset composition and limits: - Bonds, equities and real estate cover 85 percent of pension funds’ investments. - A quarter of assets include direct and indirect real estate investments; mortgages account for three percent. - Real estate-related investments (physical property and mortgages) are concentrated in Switzerland. - Foreign exposure and FX risk: - More than half of the bonds and equities held are in foreign currency. - A significant share of FX risks remains unhedged, exposing pension funds to Swiss Franc strengthening. - Shift toward riskier assets: - Pension funds increased holdings in riskier assets (mostly at the cost of bonds) during the low-interest rate period. - Average real estate exposure is 23 percent relative to an investment limit of 30 percent. - Pension funds hold around three percent in mortgages and may have indirect exposure via investment funds. - Investment returns and volatility: - Capital weighted annual net returns varied between ten percent in 2019 and minus nine percent in 2022. - The average and median performance in 2023 was around five percent. - No significant differences in net performance by fund size were observed. ### Coverage ratios, solvency, and reserves - Coverage ratios and underfunding: - The capital weighted ratio for pension funds without state guarantee was 115.6 percent in the first half of 2024. - The number of pension funds with a coverage ratio below 100 has fallen to less than one percent. - Underfunded pension funds are mainly those with a state guarantee, accounting for more than 80 percent of underfunded cases. - Recovery frameworks and burden sharing: - Distinction between minor underfunding (can be restored within five years without restructuring) and significant underfunding (requires restructuring plan). - Expected recovery period: five to seven years, but no longer than ten years. - Recovery contributions are mainly borne by employers, though the relative share borne by employees has been gradually increasing from 15 to 30 percent. - Technical rates and disclosure: - Applied technical interest rates are on average higher for pension funds with state guarantee than for private sector pension funds. - Target fluctuation reserves are on average two percent lower for those with state guarantee, implying disclosed coverage ratios may overstate relative strength. - OPSC sets an upper bound for technical interest rates: 3.19 percent for periodic tables and 2.89 percent for generation table. - Fluctuation reserves: - Fluctuation reserves range between ten and twenty percent depending on asset allocation. - Reserves were run down considerably in 2022 due to interest rate increases and only partially replenished by end 2023. ### Demographics, conversion rate, and intergenerational effects - Conversion rate and guarantees: - The statutory conversion rate has remained unchanged since 2014. - An average 1.5 percent difference between actual and statutory rates indicates most pension funds provide benefits significantly beyond the regulatory minimum. - An estimated ten percent of all pension funds are bound by the conversion rate of 6.8 percent as they only provide benefits at or close to the regulatory minimum. - The conversion rate of 6.8 percent is implicitly challenging given increasing life expectancy and low interest rates. - Compensatory measures and redistribution: - Measures to address increased technical reserves include higher employer/employee contributions, one-off payments, or partially linking benefits to capital market performance. - Pension funds providing the regulatory minimum are more likely to need contribution increases, potentially causing intergenerational redistribution. ### Box: The Unsuccessful 2024 Pillar 2 Reform - September 2024 referendum rejected reform to the Law of Occupational Benefits (LOB) by a two-thirds majority. - Reform aimed to strengthen Pillar 2 financing given increasing life expectancy and prolonged low interest rates. - Proposed changes included: - Lowering the conversion rate from 6.8 percent to 6 percent, combined with compensatory measures (coordination adjustment to 20 percent of salary, fixed permanent increase of pension for active members closer to retirement). - Lowering the threshold for mandatory occupational pension contributions to an annual salary of CHF 19,845 compared to a threshold of 22,050 (both figures related to 2024). - Adjustment to annual pension contributions for different age brackets, leading to lower contributions for older employees (relative to status quo). - Reasons cited for referendum failure included technical complexity, perceived over-compensation raising short- to medium-term costs, and redistributional concerns. - Immediate impact on the sector is low given only around 10 percent of institutions provide coverage at or near the statutory minimum; those institutions will likely need to increase contributions. - Reform failure may further push consolidation of the sector. ### Lump sums, withdrawals, and potential social consequences - Minimum and observed withdrawals: - Pension funds must accept a minimum of 25 percent in lump sum payouts but may allow higher withdrawals. - Complete withdrawals accounted for more than 40 percent in 2023; partial withdrawals add another 20 percent. - Distributional patterns: - Median lump sum is consistently higher for retirees who annuitize part of their pension funds relative to those who withdraw everything, suggesting people with lower pensions are more likely to withdraw everything. - Emigration after short work spells in Switzerland may partially explain higher full withdrawals. - Risks: - Increasing lump sum take-up may lead to frontloading of consumption and early depletion of funds, with potential future need for state support. - Small pension pots can be paid out due to high administration costs. ### Insolvency and guarantee fund - Insolvency payouts by the guarantee fund have been stable and regularly amount to less than one hundredth of a percent of pension fund liabilities. - Insolvencies have mostly affected small institutions, growing the reserve fund and reducing insolvency contributions to 0.002 percent since 2023. ### Joint and collective foundations: membership dynamics and governance - Active-to-retired member ratios: - Joint and collective foundations have almost five active members per retiree. - Single employer pension funds have around three active members per retiree. - Pension funds with several employers have a bit more than two active members per retiree. - Liquidity and competition risks: - Joint and collective foundations face pressure from competition and risks from employers switching foundations. - An employer can withdraw 100 percent of accumulated retirement of its employees from an under-funded joint or collective foundation, potentially worsening strains and prompting follow-on withdrawals. - Foundations are not required to accept a new company. - Governance concerns: - Several joint or collective foundations belong to life insurers; Article 51c LOB requires arm’s length transactions. - Supervisors should pay attention to transactions to avoid perceptions of unfair treatment or conflicts of interest. *Source: IMF staff calculation based on OECD, Swisscanto, SNB, FSO and OPSC data* ### 107. Albeit not strictly precluded, the law does not envisage on-site inspections by the ### 107. Albeit not strictly precluded, the law does not envisage on-site inspections by the ### On-site inspections and supervisory oversight - Auditing firms, paid by the supervised entity, inspect pension funds on-site. - Concerns: standardized regulatory audit reports do not provide a sufficient level of individualization to different institutions, reducing identification of idiosyncratic risks and the value of such reports to the authorities. - Recommendation: Supervisors should reduce the reliance on external regulatory auditors. - Policy requirement: Policymakers should provide supervisors with the relevant authority to perform on-site inspections. - Implementation guidance: Given the large number of institutions, on-site inspection and more intrusive supervision should be performed in a risk-based manner, i.e., focusing on risk areas and institutions with the highest probability and impact. - Note on liquid assets: "Liquid assets (as a share of total assets) have gradually fallen over the years, but data do not show significant differences across broader pensi0n fund categories. It is noticeable though that liquid asset holdings for joint foundations than for other pension funds (5.2 percent versus 4.2 percent on average across all funds." *Interconnectedness and Contagion Analysis* ### A. Methodology - Primary focus: contagion risks and vulnerabilities in the domestic interbank market. - Data and model: - Supervisory data on domestic interbank exposures. - Network analysis and default simulations based on the Espinosa-Vega and Sole (2010) approach. - Modeled system-wide impact of a hypothetical default of each bank, including first order and spillover effects from additional rounds of induced failures. - Simulation mechanics: - Computed system-wide losses from credit shocks when a counterparty defaults and funding shocks inducing liquidity shortfalls, and the combination of both. - Default defined as the depletion of excess capital, calculated as the difference between total capital and the minimum capital requirement per bank (the latter defined to include Pillar 1 & 2 requirements and SIB surcharges; excluding CCoB and CCyB buffer requirements). - Data coverage: bilateral bank exposure data available from 2014-24, enabling model application historically and at recent points in time. - Cross-border analysis: - Used BIS locational cross-border exposure data to model network effects of an assumed default of any one jurisdiction in the sample of 29 countries. - BIS data captures banks’ cross-border exposures to banks, NBFIs, governments, households, and nonfinancial firms. - Analysis limited to the 29 jurisdictions that report country-level counterparty claims. ### B. Pre-Model Analysis — domestic structure and cross-sectoral exposures - High-level findings: - Sectoral exposure data shows banks are highly exposed to OFIs, on par with the level of interbank exposures. - Domestic banks have lesser exposures to pension funds, insurance corporations, investment funds, and MMFs. - Banks’ exposures and the size of OFIs imply OFIs may present contagion risk and have implications for financial stability. - Banks’ exposure to real estate investment funds (total estimated size of 14 percent of GDP) merits more scrutiny given the systemic risk profile of the real estate market in Switzerland. - Recommendation: Authorities may consider strengthening data collection and dissemination mechanisms for OFIs to better inform contagion risk analysis and identify potential sources of vulnerability. - Cross-border linkages: - BIS data shows Switzerland’s cross-border linkages are significant. - Swiss banks have significant claims and liabilities to seven countries; the relative magnitude of these exposures is uneven. - Switzerland had a total of USD 706 billion in cross-border claims as of 2024Q3, and USD 633 billion in liabilities. ### C. Default simulations and impact and vulnerability rankings - Simulation metrics: - Impact metric: average loss to all other banks as a percentage of total excess capital. - Vulnerability metric: capital loss to a given bank induced on average by all other banks defaulting (one after another), as a percentage of the bank’s own excess capital. - Key results: - Hypothetical default of SIBs would cause the most notable domestic spillover. - Cantonal banks appear more vulnerable than other clusters. - Contagion risk of SIBs has remained elevated over time. - Default simulations identified a few banks with potential to amplify spillovers due to notable and persistent impact and vulnerability estimates. - Cross-border resilience: - Switzerland ranks 18th most vulnerable to cross-border spillovers in the sample of 29 jurisdictions. - Switzerland is particularly susceptible to cross-border contagion risks from the U.S., Japan, France, Germany, Italy, the Netherlands, and the UK. - Switzerland ranks 10th on its impact metric for induced cross-border failures, indicating contagion risk from Switzerland is low. *CLIMATE RISK ANALYSIS* ### A. Pre-Model Analysis — flood risk context and exposure - Physical risk context: - Climate warming, population growth, and land use intensification could increase frequency and intensity of flood events. - Evidence: with an increase in Global Warming Levels (GWL), more intense heavy rainfalls will lead to more localized flooding and surface runoff; large-scale floods could become more common (NCCS, 2021). - As population grows and landscape use intensifies, extreme flood events effects could double under a 2-degree GWL (Munz et al., 2024). - Banks have significant exposure to mortgages and non-amortization is a significant part of the portfolio; increased climate-related damages could impact households and banks. - Data and exposure layers: - Used residential building locations, household microdata from SILC and HBS, and banks’ canton-level mortgage portfolio distributions. - Residential building locations from the FSO calibrate flood-related indicators; household microdata calibrates damage to real estate by major regions. - Geographical mortgage exposure distribution varies by bank, with higher geographical concentration among cantonal and regional banks. - Hazard indicators: - Current conditions indicator: percentage of residential buildings in flood zones for each canton using FOEN flood data and FSO building data, translating return periods into an average annual percentage of buildings at risk. - Future conditions indicator: relative change of average annual flood depth using WRI flood maps and residential building data. - Event-related impacts: flood dynamic scenarios from Munz et al. (2023), covering nine flood events, defined as probability of being affected and damage rate at cantonal and major region levels. - Hazard data used to assess acute risk impact due to scenarios depicting joint location effects. - Exposure heterogeneity: - Banks’ exposure to different physical hazards varies by bank and indicator. - Cantonal and regional banks show the highest indicators in the worst-case scenario. - Banking system aggregate mortgage exposure concentrated in some cantons, but those cantons are not uniformly the ones with higher current risk or expected future increases. - For plausible flood events, the worst-case scenario shows significantly greater effects across all banks and different institution types; this scenario serves as the acute risk impact assessment on LGDs. ### B. Methodology — scenario design and bank impact estimation - Scenario design: - Climate risk scenario analysis focused on banks’ mortgage portfolios and physical risk (flood risk) under RCP 8.5 conditions. - Impact channels: acute physical damage to real estate (loss-given default) and chronic effects via increased insurance premiums raising default probabilities. - Considered worst-case scenario from nine flood scenarios under current and 2-degree GWL (Munz et al., 2024) and increase in annual expected damage from river floods for increasing insurance premiums. - Bank-specific implementation: - Impacts are bank-specific, depending on geographical footprint at major region level, leveraging microdata and the mortgage portfolio model of the bank solvency stress test. - Estimation of property value impacts and insurance expense increases: - Property value impact estimation used household micro-data, probability of being affected, and damage rate from the worst-case flood event and a 2-degree GWL. - Sampling over household combinations to obtain a property value impact distribution for each major region, using the average in the tail of this distribution (top 5 percent). - Increase in insurance expenses used a single value of average annual flood damages corresponding to the median under RCP 8.5 for mid-century for Switzerland from NGFS, which exceeded relative changes estimated with WRI. - The average annual flood damages relative change was applied to each household’s building insurance expenses; the weighted average increase in building insurance expenses for households with mortgages serves as input to the model. - Context on building insurance (Box 4): - Building insurance in Switzerland provided through two systems: - 19 cantons: local public sector insurers, Kantonale Gebäudeversicherungen (KGVs), not-for-profit. - Remaining cantons (Geneve, Uri, Schwyz, Ticino, Appenzell Innerrhoden, Valais, and Obwalden — "GUSTAVO cantons"): private insurance firms; insurance is compulsory in cantons with public insurers and in three of the GUSTAVO cantons; private insurers supervised by FINMA. - Both systems operate under "double solidarity": - Solidarity among insureds: KGVs charge the same price (as a percentage of insured value) independently of property location risk; private system charges the same price across GUSTAVO cantons reflecting a seven-canton risk profile. - Solidarity among insurers: public system uses an inter-cantonal fund for excess losses jointly financed by KGVs and managed by the Intercantonal Reinsurance Association (IRV); private system has a compensation fund among private insurers. - KGVs also cover response and prevention: financing training and equipment of response brigades, intervening in building permission stage, promoting and financing damage-reducing improvements, collaborating with risk mapping, and defining projects to reduce risk and damages. *Source: IMF staff analysis as presented in the provided content.* ### 120. Household PDs are affected by rising insurance costs, while LGDs are affected through ### 1cheea2025011-source-pdf - 120. Household PDs are affected by rising insurance costs, while LGDs are affected through ### Climate risk channels to household credit risk (PDs and LGDs) - The structural micro-macro simulation model (IDHBS+) was used to estimate the impact of: - physical damage on properties (affecting LGDs), and - the increase in insurance expenses (affecting PDs). - PD channel: - The model maps the increase in insurance expenses into rising household consumption (insurance expenses are a component of consumption). - By assumption, the shift in expenses is the same for all households. - The change in expenses leads to a change in PD. - LGD channel: - The model translates physical damage into a loss of real estate values, by major region. - LGD effects were computed bank-specifically by taking each bank’s regional mortgage exposure profile and the associated regionally differentiated physical risk impact into account. - PD and LGD impacts are combined to obtain a total capitalization impact for all banks. ### Results: physical climate risk impact on banks (flood scenarios and insurance) - An extreme flood scenario notably reduces real estate values, with the LGD channel adversely affecting bank capital more than the PD channel (via increasing premiums). - Impacts are significantly higher in the Worst case and 2-degree GWL scenarios. - The most pronounced impacts were estimated for cantonal and regional banks, explained by: - more notable mortgage exposure, and - their provision to flood-exposed regions. - Under the assumption that insurance coverage of losses will prevail in the future, the LGD impacts are nullified. - Insurance-transfer mechanism features that limit system impact: - Swiss compulsory building insurance system redistributes risk, avoiding adverse selection and stabilizing price fluctuations via the double solidarity principle: - among insureds (same premiums within cantons), and - among insurers (inter-cantonal fund, and compensation fund in GUSTAVO cantons), - and considering prevention measures on top of this. - Estimates are limited to the specific set of flood events analyzed; trade-offs could be more pronounced due to other phenomena or regions not affected in the scenario. ### Holistic vulnerability assessment of the Swiss banking sector — Methodology - The holistic vulnerability analysis combines 20 metrics from: - initial conditions (solvency, profitability, liquidity, large exposures), and - forward-looking model results (solvency and liquidity stress tests, vulnerability metric from default simulations in interconnectedness context, a climate exposure metric). - Weights: - Initial conditions vs. forward-looking results weights are set to 25 and 75 percent, respectively. - Procedure: - All twenty metrics are used to compute indicator-specific ranks, then composite rankings are derived using the judgmentally set weights. ### Holistic vulnerability assessment — Findings (cluster-level) - Based on initial conditions: - The residual cluster of banks is the comparably weakest, followed by SIBs and regional banks. - Residual cluster weakness explained by lower initial capital ratios, weaker ROA, higher cost of risk and cost to income metrics, and lower unencumbered asset-liability coverage. - Based on forward-looking conditions: - Asset managers and SIBs appear comparably stronger, followed by regional and cantonal banks. - Asset and wealth managers’ forward-looking rankings benefit from favorable starting solvency conditions, mid-field liquidity stress test performance, and limited exposure to climate risk. - Overall composite rankings (initial + forward-looking): - Largely resemble the forward-looking ranking due to the high weight on forward-looking analysis. - SIBs’ weaker initial conditions bring their composite rank distribution closer to cantonal banks. - Additional findings: - Modest correlation between solvency and liquidity weakness at bank level. - One bank was identified as both impactful and vulnerable based on default-simulation metrics in the interconnectedness analysis. - Selected regional banks appear vulnerable on liquidity metrics but comparatively resilient on solvency metrics. ### Recommendations — Banking and system-wide risk analysis (data collection) - Authorities are recommended to collect more data in various dimensions to enhance risk analysis: - For banks: default rates at bank-portfolio level, NPL cure rates, historically and forward; the forthcoming loan-level data set (expected to be operational in 2027) should ideally allow obtaining debt metrics, e.g., DSTIs as a flow-flow debt service metric, pertaining to total debt of a debtor. - Mortgage lending: regular information on amortizing vs. non-amortizing loan contract modalities and their shares in banks’ loan books. - Banks’ interest income and related rates consistently for outstanding balances and new business at bank-portfolio level, with sufficient portfolio granularity, for all banks. - Detailed coverage of Lombard loans and all their risk parameters, especially for asset and wealth management banks and other large banks engaging in this business. - Detailed and systematic data across banks (beyond SIBs) for their trading and investment portfolios, including splits into HFT, AFS, HTM; duration metrics (residual duration, Macaulay duration for bond exposures); book and market values for HTM and AFS bond exposures; types and composition of banks’ holdings in collective investment schemes. Recommended counterparty splits: nonfinancial corporate, financial corporate, sovereign; geography: domestic, foreign, specific other countries; instruments: money market instruments, bonds, equity, etc. - More data for detailed components of banks’ fee and commission income and expenses. - Bilateral exposure data beyond banks, including all NBFI types operating in Switzerland. - More data related to CRE firms: activities, financial health, recourse to market vs. bank borrowing, extent of collateralization; and data on banks’ CRE exposures and on non-bank creditors that provide bond-based finance to CRE firms. - Develop transaction price indexes for commercial real estate (currently only ask price measures and rent prices exist). - Improve regulatory data for banks’ foreign exposures, e.g., to identify foreign CRE exposures and their risk parameters. ### Recommendations — Model development - Authorities are recommended to further develop models, specifically: - Keep developing micro data-based models for nonfinancial corporates, including as inputs to bank solvency stress testing. - Consider developing a micro data-based model for the household segment; the FSAP employed such a model using micro data from the FSO (SILC and HBS). - Develop a liquidity stress test model with all Swiss banks in scope. - Augment the solvency stress test model to include asset and wealth management banks. - Extend network analysis and contagion modeling beyond banks to include NBFIs (linked to the recommended data collection). *International Monetary Fund staff analysis based on IMF staff models, SNB, and FINMA data.* ### 131. The authorities are recommended to keep examining banks’ valuation models for real ### 1cheea2025011-source-pdf - 131. The authorities are recommended to keep examining banks’ valuation models for real ### Banking: valuation models, stress testing, and supervisory collaboration - Recommendation: FINMA is recommended to keep examining the banks’ valuation models for real estate collateral at origination, as well as the appraisal and re-evaluation processes along the lifetime of outstanding loan contracts. - Objective: ensure that appraisals be conducted in a prudent manner, not resulting in upward biases, and for valuations to be sufficiently frequent. - Recommendation: FINMA and the SNB are recommended to collaborate regarding the reciprocal benchmarking of bottom-up and top-up stress test results and the underlying models. - Bidirectional feedback and joint work would benefit both the TD and the BU analysis, for further improving the TD models and for conducting quality-assurance of the BU results. - Recommendation: It is recommended to further split the NFC segment, e.g., by separating CRE exposures. ### Mortgage LTVs, borrower-based measures, and tax incentives - Recommendation: The authorities should closely monitor the mortgage LTVs for all banks. - Rationale: Relevant in view of the recent loosening of the LTV cap in the residential investment property segment. - Policy response trigger: Should the share of mortgage loans with unduly high LTV ratios increase, the authorities should consider adequate policy responses, e.g., via binding borrower-based measures. - Recommendation reiterated from 2019 FSAP: consider abolishing the tax deductibility of mortgage interest payments. - Rationale: The current tax regime provides incentives for households to take on more mortgage debt than they otherwise would, contributing to a high level of mortgage debt and increasing household vulnerabilities. - Suggested action: A more in-depth follow-up analysis on the side of the Swiss authorities will help corroborate this recommendation further. ### SNB disclosure and macro-financial scenario transparency - Recommendation: The SNB is recommended to publish more details in its FSR, related to macro-financial scenarios and its stress test results. - Suggested content: The SNB should consider providing more quantitative information regarding the stress test scenarios and bank stress test results in its FSR, as is common practice at central banks in other jurisdictions. ### Insurance sector: liquidity stress testing, data collection, and exposures - Recommendation: FINMA is recommended to implement top-down liquidity stress testing to identify potential sources of stress for insurers. - Data needs: To support internal analyses, FINMA should collect detailed derivatives data as well as surrender and lapse information by type of insurance product. - The recently introduced detailed asset template that supports the prudent person principle provides the level of granularity needed to support the analysis of derivatives exposure. - Reporting frequency: Given the nature of liquidity risk, it is recommended to increase the reporting frequency beyond annual. - Recommendation: Reducing the dependence on ad-hoc data requests would be desirable. - Rationale: Data collections for insurers during the pandemic have shown that Swiss insurers are able to provide reliable data efficiently; supervisors should have data available on a regular, standardized basis to support regular monitoring and be less reactive. - Recommendation: FINMA should continue to closely monitor spread risks and real estate-related exposure, including any risk mitigating techniques and valuation approaches applied. - Rationale: Spread risk and property risk are the main macro-financial drivers of Swiss insurers’ solvency position. Insurers with a large property portfolio, combined with a sizable mortgage book and other real estate-related assets are particularly exposed by a real estate shock. ### Pension funds: data gaps, transparency, and supervisory tools - Finding: Data gaps and lack of transparency in a pension fund context significantly compromise market-wide analyses. - Official statistics are collected on an annual basis and are only available with considerable delay. - Market changes (example: data from 2022) can have a significant impact on pension funds’ funding position, the aftermath of which can only be analyzed several months after the end of the year. - Recommendation: Authorities should take a risk-based supervision approach, identifying clusters of entities that they monitor more regularly and more intensively. - Scope: Analyses should include both public and private pension funds, as well as a subset of single and multiple employer pension funds but in particular, joint and collective foundations due to their size and insufficient transparency. - Rationale: Given the on-going consolidation in the market, a significant market share can be captured by a comparatively small subset of institutions. - Recommendation: Further convergence of technical bases and transparency is needed, including the possibility for authorities to intervene. - Observations: Considerable variation in the choice of technical interest rates and the fluctuation reserve; weaker institutions apply more generous assumptions, implying their actual coverage ratio is worse than presented. - Need: Crucially, relevant cantonal authorities must have the instruments and authority to intervene when they identify discrepancies. - Recommendation: A formal platform to discuss market/sectoral developments and potential risks to financial stability should be established. - Participants: All authorities involved in pension supervision, i.e. cantonal supervisors, OPSC and FINMA. - Use: Findings should be fed into the discussions of the Steering Committee/Standing Committee for Financial Stability, with corresponding representation of the pension authorities at the committees, when pension issues are tabled. ### Climate risk analysis - Recommendation: The authorities are recommended to collaborate with agencies and institutions that are developing climate-related work in the country. - Suggested partners: building insurers, FOEN, and other agencies and research institutions. - Objective: develop capacity to further analyze climate risk and its implications for the financial system specifically and the Swiss economy and population at large. *Source: Excerpt from the Switzerland Financial Sector Assessment (selected recommendations and analysis).* ### Appendix II Table 1. Switzerland: Risk Assessment Matrix (Concluded) ### Appendix II Table 1. Switzerland: Risk Assessment Matrix (Concluded) ### Risk Assessment Matrix — key risks and expected impacts - Cyberthreats - Likelihood: High - Expected impact when realized: Cyberattacks on physical or digital infrastructure and service providers (including digital currency and crypto assets) or misuse of AI technologies trigger financial and economic instability. Switzerland is a leader in cross-border asset management and fintech, prone to cyberattacks. Successful attacks can lead to outages of information and communication technology systems and jeopardize the goals of availability, confidentiality, and integrity, compromising the attractiveness of the financial system. - Systemic financial instability - Likelihood: Medium - Expected impact when realized: Financial sector instability in major Swiss banks’ counterparts and political uncertainty (e.g., from elections) trigger market dislocations, with cross-border spillovers affecting weak banks and NBFIs. In the event of a severe financial crisis in major jurisdictions to which the Swiss banking system is exposed, Swiss banks could face adverse spillover effects, including through materializing credit risk, rising funding costs, and liquidity risk events. - Extreme climate events - Likelihood: Medium - Expected impact when realized: Extreme climate events driven by rising temperatures cause loss of human lives, severe damage to infrastructure, supply disruptions, lower growth, and financial instability. Switzerland is affected by climate change. Related extreme events have become more frequent over the past century, often causing significant economic losses. Such developments also bring challenges to insurance and reinsurance companies. - RAM probability convention (staff subjective assessment, next 1–3 years): - “low” = below 10 (percent) - “medium” = 10–30 (percent) - “high” = 30–50 (percent) *Italic line: Appendix II Table 1 content as provided in the source.* ### Macro‑Financial scenarios — structure and variables covered - Scenarios and horizon: Baseline and two adverse scenarios (Adverse (D) = demand shock; Adverse (S) = supply shock), with projections for 2025–2027. - Variables included (numbered in table): 1. Real GDP 2. GDP Deflator 3. Nominal GDP 4. CPI 5. Core CPI 6. Unemployment Rate 7. Wages: Nominal 8. Wages: Real 9. Interest Rates Switzerland (short-term a.av.; long-term a.av.; term spread a.av.) 10. Effective CHF: Nominal 11. Effective CHF: Real 12. Residential Property Prices: Nominal 13. Residential Property Prices: Real 14. Commercial Property Prices: Nominal 15. Commercial Property Prices: Real 16. Stock Prices: Nominal 17. Stock Prices: Real 18. Private Sector Credit: Nominal 19. Private Sector Credit: Real 20. Global Commodity Prices 21. Oil Price (Brent Crude) - Scenario outputs are provided as levels (a.f. = annual flow, 2024=100; a.av. = annual average) and YoY percent changes across 2025–2027 for baseline and adverse scenarios. ### Macro‑Financial scenarios — market risk component (selected shock specifications) - Equity prices (shocks in %): - Switzerland: -50-45 - France: -50-45 - Germany: -40-35 - Spain: -50-45 - US: -40-35 - Other Non-AE Rest of the World: -50-45 - Other Advanced Economies (AE): -40-35 - Japan: -40-35 - UK: -45-40 - Italy: -55-50 - Netherlands: -35-130 (table shows mixed ranges across scenarios) - Brent crude oil (USD) shock in %: -25-150; other commodities shock in %: -50-100 (table provides scenario ranges) - Real estate prices (shocks in %): residential and commercial property shocks vary by country and scenario (table provides ranges such as -40-30, -45-35, etc.) - Foreign exchange rates (shocks in % vs CHF): USD-CHF, EUR-CHF, GBP-CHF, JPY-CHF specified with ranges in the table (e.g., Switzerland-35-25 for some scenarios; Euro area-30-20; UK-35-25; US-35-25; Rest of the world-40-30) - Long-term interest rates (10Y) shocks (in bps) and short-term interest rates (1Y) shocks (in bps): ranges provided across jurisdictions (e.g., Switzerland, France, Germany, Spain, UK, US, Japan, Other Non-AE Rest of the World), including scenario-specific ranges (for example, long-term and short-term ranges such as -50-165; -15270; -20195 as presented in the source table). ### Market risk component II — corporate bond spread scenarios - Corporate bond spreads by rating and jurisdiction are tabulated across Adverse (D) and Adverse (S) scenarios. Sample entries (spread levels in bps by rating bucket and jurisdiction): - Switzerland: AAA 70 75 60 65 (multiple scenario rows), AA 110 115 100 110, A 275 245 240 230, BBB 310 290 280 270, BB 465 435 420 405, B and 5 percent of a bank’s total assets) - Methodology and model components: - Balance sheet model accounting for credit risk, interest rate and market risk, other P&L components, RWA; dynamic balance sheet; combination of structural and econometric components. - Credit risk: structural model (micro-macro simulation model, the IDHBS+ model) for household mortgage portfolios (PD and LGD), econometric satellite models for nonfinancial corporates, satellite models based on Moody’s KMV PDs for financial corporates using BMA methodology. - Interest income/expense: econometric pass-through equations. - Net fees/other income: bank panel econometric models using BMA. - Market risk: modified duration approach for bond valuation; partial account for hedging. - STA risk weights constant; IRB risk weights modeled dynamically. - Stress test horizon: 3 years: 2025-2027 - Analyses: baseline (WEO) and two adverse scenarios (demand shock/disinflationary; supply shock/inflationary); sensitivity analysis includes mark-to-market for HTM bonds and solvency-to-funding-cost feedback on vs. off. - Capital buffers and reporting: Pillar 1 plus prudential buffers (Pillar 2), CCoB, CCyB, SIB surcharges considered; CCoB and CCyB allowed to be “consumed” under adverse scenarios; reporting of system-wide capital evolution/depletion and capital shortfalls, aggregated contributions by bank clusters. ### Banking sector liquidity stress test — key design features - Institutional perimeter and data: same institutions and market share as solvency test; regulatory data based on Basel III standardized liquidity monitoring tools; cut-off date end-December 2024. - Methodology and channels: cash flow-based liquidity stress test with liquidity-solvency feedback and add-back mechanism for secured funding; linkage to market risk via equity and bond revaluation in liquidity model. - Monitoring metrics: - Liquidity coverage ratio (LCR) in CHF (requirement) and in significant currencies (EUR, GBP, JPY, USD) (monitoring metric) - Net Stable Funding Ratio (reporting requirement) - Concentration of funding (monitoring metric) - Risks and buffers: funding risk, rollover risk, market liquidity risk; buffer = stock of liquid assets. - Tail shock calibration: runoff shock informed by LCR parameterization and historical experience; revaluation of bond holdings in counterbalancing capacity in line with market risk shocks from solvency stress test. - Regulatory standards: Basel III full implementation for the LCR ratio at 100 percent; counterbalancing capacity above net cash outflows under stress scenario. - Reporting outputs: changes in average liquidity position and counterbalancing capacity by scenario; distribution of banks’ liquidity position by scenario; number of banks with counterbalancing capacity below net cash outflows; banks’ post-shock net liquidity position; liquidity shortfall in terms of banking system total liabilities. ### Insurance sector solvency stress test — design and shocks - Institutional perimeter and data: six insurance groups covering two thirds of domestic life and non-life premia; companies’ own data and FINMA regulatory reporting; reference date regulatory reporting December 31, 2024 (and December 31, 2023 for sensitivity analysis (cyber) in BU by insurers / TD by IMF columns). - Methodology and horizon: investment assets valued by market value changes after price shocks; sensitivity analysis on available capital and solvency position; time horizon includes instantaneous shock and 3-year projection. - Tail shock (Adverse scenario) summary (selected elements repeated from insurance sector specification): - Risk-free interest rates: 107 bps (1y CHF), 337 bps (10y CHF), 328 bps (1y EUR), 258 bps (10y EUR), 478 bps (1y USD), 502 bps (10y USD), 479 bps (1y GBP), 501 bps (10y GBP) - Sovereign bond spreads: 30 bps domestic, 100 bps low spread EEA countries, 200 bps other EEA countries, 95 bps United Kingdom, 65 bps United States, 30 bps Japan, 100 bps other advanced economies, 200 bps emerging and developing countries - Stock prices: -45 percent domestic, -40 percent other advanced economies, -50 percent emerging and developing economies - Property prices: domestic -25 percent (residential) and -30 percent (commercial); foreign -30 percent (residential) and -35 percent (commercial) - Corporate bond spreads: ranges between 60 bps (AAA) and up to 570 bps (B and lower financials) noted in the insurance sector specification - Mortgage default spread increase by 150 basis points - Appreciation of CHF against major currencies (EUR, USD, JPY: 15%, GBP: 20%) - Sensitivity analyses: outage of cloud service provider; Petya/WannaCry-type ransomware attack; other specified sensitivities. - Regulatory/accounting standards: Swiss Solvency Test; National GAAP, IFRS, US GAAP. - Reporting outputs: impact on solvency ratios; contribution of individual shocks; impact of reactive management actions; dispersion measures of solvency ratios and assets over liabilities. ### Interconnectedness analysis — approach and outputs - Institutions involved: Swiss banks and counterparties in other countries (BIS cross-country exposures), using Swiss banks’ bilateral interbank exposure data and SWIFT cross-country transaction flow data. - Data and starting position: Swiss domestic sectoral exposures; BIS cross-country exposures; additional data for assets, RWAs, capital; largest cross-exposures; SWIFT transaction flow data. - Methodology: domestic cross-sectoral exposure analysis (net lenders/net borrowers); network metrics; default contagion simulations using Espinosa‑Vega & Solé (EVS 2010) model based on BIS cross-country exposure data and Swiss banks’ interbank exposures; systematic shock simulations (each node fails sequentially) and triggers informed by solvency/liquidity stress test results. - Buffers and outputs: banks’ capital and liquid asset buffers; outputs include network metrics and visualizations, EVS model results with bank and country level impact and vulnerability rankings, capital losses, and emphasis on entities ranking high in both impact and vulnerability. ### Climate risk analysis — scope and methodology - Institutions included: same institutions as in the banking sector stress testing, covering mortgage portfolios (dependent on availability of SNB data on geographical distribution). - Data and cut-off date: public and supervisory data; cut-off date December 2024; public or authorities’ data on climate projections and insurance conditions. - Methodology: structural model (micro-macro simulation model) for household mortgage portfolios, for both PD and LGD components, rooted in micro/household survey data, geographically differentiated, linked to climate projections, and considering insurance conditions. *Italic line: Content drawn directly from the source PDF tables and appendices provided.* ### 4. Scenarios ### 4. Scenarios (content unit) ### Scenarios and time horizon - Baseline scenario: same as in the banking sector stress testing (current climate conditions) - At least one adverse scenario: based on future climate conditions from IPCC (depending on the availability of AR5 or AR6 and heterogeneity of future climate conditions) - Time horizon: 3 years: 2025-2027, and considering future climate conditions for 2050 and 2100 ### Risks/factors assessed - The impact of chronic and acute climate conditions on households and its effects on PD and LGD components ### Output presentation - Delta PDs and delta LGDs - Individual banks’ capital ratio impacts --- ### Appendix V — Bank Credit Risk Model Details (For Segments Other Than Residential Real Estate) ### Modeling approach and data coverage - A Bayesian Model Averaging (BMA) methodology was employed to estimate the corporate PD models. The methodology explicitly accounts for model uncertainty and optionally involves coefficient sign constraints. - Left hand-side variables: sectoral PDs sourced from Moody’s Credit Edge (1-year expected default frequencies, EDFs). - Four sub-segments defined: - NFC (307 Swiss firms) - CRE (30 firms) - Banks (32 firms) - NBFIs (63 firms) ### Model specification and constraints - Sign constraints imposed on long-run multipliers of predictors to obtain economically meaningful coefficient signs and scenario conditional forecasts. - Allowed predictors across four segments: GDP growth, the unemployment rate, and interest rates. - Residential property prices allowed for CRE and bank segments (relevant for the LGD channel for Swiss banks’ sizeable mortgage portfolios). - Leverage allowed to drive PDs in the four sub-segments, using respective own leverage metrics. Additional scenario paths designed for leverage (Figure E2). - Pandemic period excluded from estimation for all models (to avoid biasing macro-financial sensitivities toward zero). - Maximum number of right hand-side variables per equation in the model space: 5. - Allowed lags: beyond contemporaneous inclusion, two additional lags were allowed (optimized per equation). - Left hand-side transformation: logit transformation of the PDs considered to guarantee projected PDs remain in the 0-1 interval. - For NFC segment: different models explored—debt weighted vs. median PDs and including vs. excluding stock price variable. When included, stock price was a dominant driver in adverse scenarios. ### Model variants, fit, and diagnostics - NFC scenario conditional forecasts computed as an average of three model variants’ conditional forecasts. - Historical fit (logit) R-squares reported: - NFC—Debt-Weighted PD, Including Stock Prices: R-square = 0.85 - NFC—Debt-Weighted PD, Without Stock Prices: R-square = 0.34 - NFC—Median PD, Including Stock Prices: R-square = 0.98 - CRE: R-square = 0.59 - Banks: R-square = 0.85 - NBFIs: R-square = 0.47 - Residual diagnostics (difference of PD fit and PD observed, both in logit) Durbin Watson (DW) statistics: - NFC—Debt-Weighted PD, Including Stock Prices: DW = 2.2 - NFC—Debt-Weighted PD, Without Stock Prices: DW = 0.4 - NFC—Median PD, Including Stock Prices: DW = 2.2 - CRE: DW = 1.4 - Banks: DW = 1.6 - NBFIs: DW = 0.4 ### Predictor inclusion mapping (high-level) - Predictor pool included transformations of: - Real GDP growth (QoQRGDPQ and YoYRGDPY) - Unemployment rate (LevelUR1, Diff QoQURD, Diff YoYURD) - Short-term interest (LevelSTR, Diff QoQSTRD, Diff YoYSTRD) - Long-term interest rate (LevelLTR, Diff QoQLTRD, Diff YoYLTRD) - Policy rate (LevelPOL, Diff QoQPOLD, Diff YoYPOLD) - Real effective exchange rate (LevelREER1000, QoQREERQ1000, YoYREERY1000) - Residential property prices (QoQNRPPQ, YoYNRPPY) - Commercial property prices (QoQNCPPQ, YoYNCPPY) - Stock prices (QoQSTOQ, YoYSTOY) - Real oil prices (QoQROIL CHF Q1000, YoYROIL CHF Y1000) - Corporate bond spread (CBS) - Segment dummies/metrics (NFC, CRE, BA, NBFI) - Inclusion settings and sign constraints summarized in Appendix V. Table 1 (symbolic mapping: +1 = allow inclusion and set positive LRM sign constraint; -1 = allow inclusion and set negative LRM sign constraint; 0 = do not include). --- ### Appendix VI — Household Sector Model ### Model overview and data sources - A micro data-based structural household sector model (IDHBS framework) was set up for Switzerland based on: - Statistics on Income and Living Conditions (SILC), including wealth-specific variables module from 2022 - Household Budget Survey (HBS) 2015-2017 - SILC (2022) contained 8,900 households and 19,229 household members. - HBS contained 9,955 households. - The model draws most data from SILC (2022) with wealth-specific variables; HBS provided complementary information on household living expenses. - An additional module containing wealth data was included in the SILC dataset. ### Model mechanics and default definition - The model simulates household income, expenses, and implied balance sheet evolution (structural forward simulation). - Household balance sheet mechanics and default definition summarized in Figure F2. - For Switzerland, the default definition excluding a role for house prices was considered, given the full recourse environment and limited incentives for strategic default when collateral values fall short of outstanding debt. ### Employment model within the household framework - Logistic model for individuals’ employment status estimated based on Swiss SILC data. - Logistic model estimates based on data for 11,438 individuals (status: employed (=1) or unemployed (=2)). - Seven regional indicators included (Lake Geneva region as reference). - Model performance metrics: AUROC of 0.67 and GINI of 0.34. - Employment status simulated stochastically with an intercept shift to match target aggregate unemployment rates provided in scenarios. ### Key model parameters (selected) - Replacement rate: 72% (OECD) — Unemployment benefit over last employment income net of tax - Maximum unemployment benefit per month in CHF: 7,970 (Swiss State Secretariat for Econ. Affairs (SECO)) - Variable rate share of total household debt: 20% (SNB and own calculations) - Pass-through from policy to deposit rate (up): 0.25 (SNB and own estimates) - Pass-through from policy to deposit rate (down): 0.35 (SNB and own estimates) - PD anchor for mortgage debt: 0.5% (SNB/FINMA) - PD anchor for other debt: 1.5% (Own calculations) - LGD anchor for mortgage debt: 10% (SNB/FINMA) - Cure rate (for LGD module): 20% (Data from banks; Assumption) - Interest rate for "other debt" in 2022: 4.7% (Informed by historical interest rate data) ### Data treatments and imputations - Household expenses not included in SILC were imputed from HBS by computing living expense ratios (expenses as percentage of household gross income). Three groups considered: - Owners without mortgages - Owners with mortgages - Renters (39 percent of the households) - For owners with mortgages and renters: median living expenses ratio used in each gross income decile. - For owners without mortgages: quintiles used due to limited sample size. - Fire insurance and other building damage premiums identified for owners and used to increase insurance expenses for climate risk analysis. - Household member economic status rules for missing data: - If only unemployment benefits reported → set as unemployed; employment income inferred via inverse of replacement rate. - If only employment income reported → set as employed; unemployment benefit inferred using replacement rate. - If both reported → both converted consistently (methods (i) and (ii) applied). - Household assets and liabilities included: property value, bank accounts, bonds, stocks, mortgage and non-mortgage liabilities; main residence mortgage flows available from SILC. - Most households owning main or other residence do so through a mortgage; less than 30 percent of households have just non-mortgage debt. - Percentage owners of main residence from SILC (with and without a mortgage) and non-renters from HBS aligned: 38 and 39 percent, respectively. - Property values: when non-reported, value type defined and values updated using the residential property price index. - For second residence mortgage flows: - If household has mortgages in main residence: second residence flow ratios set equal to main residence interest and principal flow ratios. - If household has no mortgage in main residence: imputed second residence flows set to the 75th percentile of interest and principal flow ratio in each income quartile from households with mortgages in main residence. - Non-mortgage debt flows: constructed using portfolio maturity and an estimated interest rate to derive principal and interest flows. --- *Source: 4. Scenarios; Appendix V. Bank Credit Risk Model Details; Appendix VI. Household Sector Model — extracted from the provided IMF Switzerland material.* ### 9. The model’s debt data reveals differences between amortizing and non-amortizing ### 9. The model’s debt data reveals differences between amortizing and non-amortizing ### Debt dataset construction and alignment with bank data - Debt stock and flows were arranged into mortgage and non-mortgage, from which the principal repayment variable defines whether a loan is amortizing. - The household input data aligns with banks new owner-occupied mortgages from 2017 to 2024, showing similar levels of disposable income and LTVs (Figure F2). ### Key empirical differences between amortizing and non-amortizing loans (model input) - Income is higher in amortizing mortgages. - LTVs are lower in the non-amortizing (aligned with the “industry standard” of non-amortizing loans below 66 percent LTV). - Interest rates are higher in amortizing loans. - DSTI is higher for amortizing mortgages. - Sources used: SILC (2022) and IMF staff calculations. *Appendix VI. Table 5 provides related household mortgage characteristics including:* - A. Annual Disposable Income for Households with Mortgages (CHF thousands) - B. LTV Ratios for Household Mortgage Debt (CHF thousands) - C. Interest Rate for Mortgage Debt (percent) - D. DSTI Ratios for Mortgage Debt (percent; mortgage debt service over disposable income) - E. Income for Newly Granted Mortgages for Household Owner-Occupied Real Estate (CHF thousands, new mortgages 2017-2024) - F. LTV Ratios for Newly Granted Mortgages for Household Owner-Occupied Real Estate (percent, new mortgages 2017-2024) ### Bank interest income and cost of funding models (Appendix VII) - Panel econometric models project interest income and expenses for individual banks in the bank solvency stress test. - Models were set up for each bank cluster; SIBs, asset managers, and residual bank clusters were further split into domestically active and cross-border active banks sub-clusters. - For foreign bank sub-clusters, dependence on certain foreign interest rates was allowed in addition to domestic drivers. - Cost of funding (COF) models allowed an asymmetric response to changes in policy and market rates depending on direction of change. - Feedback from solvency ratios of banks to their cost of funding was allowed. - Interest income models capture pass-through from policy rates and banks’ own cost of funding; interest income rates defined as banks’ trailing annual sum of interest income over annual trailing average interest bearing assets (models based on outstanding stocks, not rates on new lending); models include an autoregressive lag to capture rate persistence. Key takeaways from COF and IIR model estimations: - COF models: - Systematic evidence for an asymmetric response of banks’ cost of funding to up- vs. downward moves in policy and market rates. - An economically non-negligible feedback from bank capital ratios to their COF was found. - The fit of the COF models is reasonable despite the models having no autoregressive lags. - IIR models: - An asymmetric response of IIRs to changes in policy and market rates was found, although the asymmetry is not as pronounced as for the COF models. - Macroeconomic control variables were not found to be too relevant. Model data, estimation universe, and presentation: - Raw data underlying the interest income and expense models are shown in Figure G1; model coefficient estimates in Figure G2; projections for COF and IIR at cluster aggregate level in Figs. G3 and G4. - Notes: Coefficient estimates obtained from bank cluster specific panel models, estimated using EGLS, based on quarterly data for 92 banks in total across clusters, covering the 2000Q3-2024Q4 period. - Figures include: - Appendix VII. Figure 1: Interest Income Rates (IIRs) and Cost of Funding (CoF) — data for 92 banks, 2000Q3-24Q4; cluster aggregates and interquartile ranges. - Appendix VII. Figure 2: Cost of Funding and Interest Income Models — coefficients. ### Projections (Appendix VII figures) - COF Projections presented at cluster level for: - All Banks; SIBs; Cantonal Banks; Asset and Wealth Managers; Regional Banks; Other Banks. - IIR Projections presented at cluster level for the same clusters. - Presentation notes: - Bars represent medians across underlying banks in respective clusters. - Error bars extend to the 10th and 90th percentiles. - Vertical axes set to a common range to support cross-cluster comparison. ### Bank fee and commission income (FCI) models (Appendix VIII) - Historical data compiled in four categories: FC expense and three FC income items (AUM, credit granting, residual). - Credit and residual sources contribute notably less to total FCI than banks’ AUM; modeling priority assigned to FCI from AUM. Modeling approach: - FX expense ratios relative to total assets were held constant. - FCI from AUM modeled using a panel BMA methodology to model ratios of FC income from AUM relative to AUM as a function of macrofinancial conditions. - FC income for credit and residual business (both relative to total assets) held constant under the baseline and follow a predefined percentile path under the two adverse scenarios. - BMA model settings for the FCI for AUM are shown in Table H1 (predictor pool composition and sign constraints). Appendix VIII model diagnostics and outputs: - Panel BMA model results collected in Figures H1-H5 for five bank clusters; percentile paths for FCI ratios for credit and other business in Figure H6; projections of various ratios from three FCI categories (medians over banks per cluster) in Figures H7-H9. - Example model fit statistics: - Appendix VIII. Figure 1 (SIBs): Historical Fit R-square = 0.98; DW = 0.82 (pandemic period 2020-21 included in reported R-squares and DW, while estimation excluded this period). - Appendix VIII. Figure 2 (Cantonal Banks): R-square = 0.80; DW = 0.6. - Appendix VIII. Figure 3 (Asset and Wealth Management Banks): R-square = 0.94; DW = 0.9. - Appendix VIII. Figure 4 (Regional Banks): R-square = 0.97; DW = 0.7. - Appendix VIII. Figure 5 (Other Banks): R-square = 0.11; DW = 0.2. Projection notes: - FCI ratios defined as 4-quarter trailing sums of FCI from AUM over 4-quarter trailing averages of AUM, at bank level; medians across underlying banks shown. - FCI from sources other than AUM and credit business defined as 4-quarter trailing sums of FCI from those sources over 4-quarter trailing averages of total assets, at bank level; medians across banks shown. - FCI from credit business defined as 4-quarter trailing sums of FCI from credit business over 4-quarter trailing averages of total assets, at bank level; medians across banks shown. ### Climate risk analysis (Appendix IX) - Hazard indicators weight residential buildings to represent exposure of interest (mortgages): each residential building hectare receives hazard parameters; indicator is weighted average of parameters using number of buildings (Figure I1). - For WRI, a similar process applies; estimates from Munz et al. (2023) are already in terms of buildings and did not require additional treatment. Flood scenarios and exposure: - Flood dynamics scenarios from Munz et al. (2023) consider physically plausible scenarios for extreme riverine flood events and quantify socioeconomic impacts. - Nine flood event scenarios expressed in terms of probability of being affected and damage rate at cantonal and regional levels: - Cantonal level used to assess banks’ cantonal exposure. - Regional level used to calibrate impact of damage on real estate. - Appendix IX. Table 1: Flood Scenarios (Munz et al. 2023). - Appendix IX. Figure 2: Flood Scenarios by Canton Distribution (percent for probability affected and damage rate). Banking system exposure visuals: - Appendix IX. Figure 3: Mortgages Exposure to Floods Actual Conditions (flood: in average annual percentage at risk; loans: percent of total loans 2015-2024). - Appendix IX. Figure 4: Mortgages Exposure to Floods Future Conditions — RCP4.5 and RCP 8.5 (flood: relative change average annual flood depth; loans: percent of total loans 2015-2024). - Appendix IX. Figure 5: Bank Exposure—Actual and Future Conditions (Banks Floods Actual Conditions: meters weighted average with banks mortgages; Banks Relative Change Floods Future Conditions: relative change weighted average with banks mortgages). - Appendix IX. Figure 6: Floods Average Annual Damage—Future Conditions (Average Annual Damage Relative Change RCP 4.5 (percent); Average Annual Damage Relative Change RCP 8.5 (percent)). Sources and data attribution - Sources referenced across Appendices: SNB, FINMA, SILC (2022), FOEN, FSO, swisstopo, WRI, Huizinga et al. (2017), Munz et al. (2023), and IMF staff calculations. *Source: 1cheea2025011-source-pdf - 9. The model’s debt data reveals differences between amortizing and non-amortizing (from the provided PDF content).* --- _Source: https://www.imf.org/-/media/files/publications/cr/2025/english/1cheea2025011-source-pdf.pdf_